ipfs-cluster/cmd/ipfs-cluster-service/daemon.go

298 lines
7.4 KiB
Go
Raw Normal View History

Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
package main
import (
"context"
"errors"
"os"
"os/signal"
"syscall"
"time"
host "github.com/libp2p/go-libp2p-host"
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
"github.com/urfave/cli"
ipfscluster "github.com/ipfs/ipfs-cluster"
"github.com/ipfs/ipfs-cluster/allocator/ascendalloc"
"github.com/ipfs/ipfs-cluster/allocator/descendalloc"
"github.com/ipfs/ipfs-cluster/api/ipfsproxy"
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
"github.com/ipfs/ipfs-cluster/api/rest"
"github.com/ipfs/ipfs-cluster/consensus/raft"
"github.com/ipfs/ipfs-cluster/informer/disk"
"github.com/ipfs/ipfs-cluster/informer/numpin"
"github.com/ipfs/ipfs-cluster/ipfsconn/ipfshttp"
"github.com/ipfs/ipfs-cluster/monitor/basic"
"github.com/ipfs/ipfs-cluster/monitor/pubsubmon"
"github.com/ipfs/ipfs-cluster/observations"
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
"github.com/ipfs/ipfs-cluster/pintracker/maptracker"
"github.com/ipfs/ipfs-cluster/pintracker/stateless"
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
"github.com/ipfs/ipfs-cluster/pstoremgr"
"github.com/ipfs/ipfs-cluster/state/mapstate"
ma "github.com/multiformats/go-multiaddr"
)
func parseBootstraps(flagVal []string) (bootstraps []ma.Multiaddr) {
for _, a := range flagVal {
bAddr, err := ma.NewMultiaddr(a)
checkErr("error parsing bootstrap multiaddress (%s)", err, a)
bootstraps = append(bootstraps, bAddr)
}
return
}
// Runs the cluster peer
func daemon(c *cli.Context) error {
logger.Info("Initializing. For verbose output run with \"-l debug\". Please wait...")
ctx, cancel := context.WithCancel(context.Background())
defer cancel()
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
// Load all the configurations
cfgMgr, cfgs := makeConfigs()
// Run any migrations
if c.Bool("upgrade") {
err := upgrade(ctx)
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
if err != errNoSnapshot {
checkErr("upgrading state", err)
} // otherwise continue
}
bootstraps := parseBootstraps(c.StringSlice("bootstrap"))
// Execution lock
err := locker.lock()
checkErr("acquiring execution lock", err)
defer locker.tryUnlock()
// Load all the configurations
// always wait for configuration to be saved
defer cfgMgr.Shutdown()
err = cfgMgr.LoadJSONFromFile(configPath)
checkErr("loading configuration", err)
if c.Bool("stats") {
cfgs.metricsCfg.EnableStats = true
}
cfgs = propagateTracingConfig(cfgs, c.Bool("tracing"))
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
// Cleanup state if bootstrapping
raftStaging := false
if len(bootstraps) > 0 {
cleanupState(cfgs.consensusCfg)
raftStaging = true
}
if c.Bool("leave") {
cfgs.clusterCfg.LeaveOnShutdown = true
}
cluster, err := createCluster(ctx, c, cfgs, raftStaging)
checkErr("starting cluster", err)
// noop if no bootstraps
// if bootstrapping fails, consensus will never be ready
// and timeout. So this can happen in background and we
// avoid worrying about error handling here (since Cluster
// will realize).
go bootstrap(ctx, cluster, bootstraps)
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
return handleSignals(ctx, cluster)
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
}
func createCluster(
ctx context.Context,
c *cli.Context,
cfgs *cfgs,
raftStaging bool,
) (*ipfscluster.Cluster, error) {
err := observations.SetupMetrics(cfgs.metricsCfg)
checkErr("setting up Metrics", err)
tracer, err := observations.SetupTracing(cfgs.tracingCfg)
checkErr("setting up Tracing", err)
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
host, err := ipfscluster.NewClusterHost(ctx, cfgs.clusterCfg)
checkErr("creating libP2P Host", err)
peerstoreMgr := pstoremgr.New(host, cfgs.clusterCfg.GetPeerstorePath())
peerstoreMgr.ImportPeersFromPeerstore(false)
api, err := rest.NewAPIWithHost(ctx, cfgs.apiCfg, host)
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
checkErr("creating REST API component", err)
proxy, err := ipfsproxy.New(cfgs.ipfsproxyCfg)
checkErr("creating IPFS Proxy component", err)
apis := []ipfscluster.API{api, proxy}
connector, err := ipfshttp.NewConnector(cfgs.ipfshttpCfg)
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
checkErr("creating IPFS Connector component", err)
state := mapstate.NewMapState()
err = validateVersion(ctx, cfgs.clusterCfg, cfgs.consensusCfg)
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
checkErr("validating version", err)
raftcon, err := raft.NewConsensus(
host,
cfgs.consensusCfg,
state,
raftStaging,
)
checkErr("creating consensus component", err)
tracker := setupPinTracker(c.String("pintracker"), host, cfgs.maptrackerCfg, cfgs.statelessTrackerCfg, cfgs.clusterCfg.Peername)
mon := setupMonitor(c.String("monitor"), host, cfgs.monCfg, cfgs.pubsubmonCfg)
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
informer, alloc := setupAllocation(c.String("alloc"), cfgs.diskInfCfg, cfgs.numpinInfCfg)
ipfscluster.ReadyTimeout = cfgs.consensusCfg.WaitForLeaderTimeout + 5*time.Second
return ipfscluster.NewCluster(
host,
cfgs.clusterCfg,
raftcon,
apis,
connector,
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
state,
tracker,
mon,
alloc,
informer,
tracer,
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
)
}
// bootstrap will bootstrap this peer to one of the bootstrap addresses
// if there are any.
func bootstrap(ctx context.Context, cluster *ipfscluster.Cluster, bootstraps []ma.Multiaddr) {
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
for _, bstrap := range bootstraps {
logger.Infof("Bootstrapping to %s", bstrap)
err := cluster.Join(ctx, bstrap)
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
if err != nil {
logger.Errorf("bootstrap to %s failed: %s", bstrap, err)
}
}
}
func handleSignals(ctx context.Context, cluster *ipfscluster.Cluster) error {
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
signalChan := make(chan os.Signal, 20)
signal.Notify(
signalChan,
syscall.SIGINT,
syscall.SIGTERM,
syscall.SIGHUP,
)
var ctrlcCount int
for {
select {
case <-signalChan:
ctrlcCount++
handleCtrlC(ctx, cluster, ctrlcCount)
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
case <-cluster.Done():
return nil
}
}
}
func handleCtrlC(ctx context.Context, cluster *ipfscluster.Cluster, ctrlcCount int) {
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
switch ctrlcCount {
case 1:
go func() {
err := cluster.Shutdown(ctx)
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
checkErr("shutting down cluster", err)
}()
case 2:
out(`
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
Shutdown is taking too long! Press Ctrl-c again to manually kill cluster.
Note that this may corrupt the local cluster state.
!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
`)
case 3:
out("exiting cluster NOW")
locker.tryUnlock()
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
os.Exit(-1)
}
}
func setupAllocation(
name string,
Feat: emancipate Consensus from the Cluster component This commit promotes the Consensus component (and Raft) to become a fully independent thing like other components, passed to NewCluster during initialization. Cluster (main component) no longer creates the consensus layer internally. This has triggered a number of breaking changes that I will explain below. Motivation: Future work will require the possibility of running Cluster with a consensus layer that is not Raft. The "consensus" layer is in charge of maintaining two things: * The current cluster peerset, as required by the implementation * The current cluster pinset (shared state) While the pinset maintenance has always been in the consensus layer, the peerset maintenance was handled by the main component (starting by the "peers" key in the configuration) AND the Raft component (internally) and this generated lots of confusion: if the user edited the peers in the configuration they would be greeted with an error. The bootstrap process (adding a peer to an existing cluster) and configuration key also complicated many things, since the main component did it, but only when the consensus was initialized and in single peer mode. In all this we also mixed the peerstore (list of peer addresses in the libp2p host) with the peerset, when they need not to be linked. By initializing the consensus layer before calling NewCluster, all the difficulties in maintaining the current implementation in the same way have come to light. Thus, the following changes have been introduced: * Remove "peers" and "bootstrap" keys from the configuration: we no longer edit or save the configuration files. This was a very bad practice, requiring write permissions by the process to the file containing the private key and additionally made things like Puppet deployments of cluster difficult as configuration would mutate from its initial version. Needless to say all the maintenance associated to making sure peers and bootstrap had correct values when peers are bootstrapped or removed. A loud and detailed error message has been added when staring cluster with an old config, along with instructions on how to move forward. * Introduce a PeerstoreFile ("peerstore") which stores peer addresses: in ipfs, the peerstore is not persisted because it can be re-built from the network bootstrappers and the DHT. Cluster should probably also allow discoverability of peers addresses (when not bootstrapping, as in that case we have it), but in the meantime, we will read and persist the peerstore addresses for cluster peers in this file, different from the configuration. Note that dns multiaddresses are now fully supported and no IPs are saved when we have DNS multiaddresses for a peer. * The former "peer_manager" code is now a pstoremgr module, providing utilities to parse, add, list and generally maintain the libp2p host peerstore, including operations on the PeerstoreFile. This "pstoremgr" can now also be extended to perform address autodiscovery and other things indepedently from Cluster. * Create and initialize Raft outside of the main Cluster component: since we can now launch Raft independently from Cluster, we have more degrees of freedom. A new "staging" option when creating the object allows a raft peer to be launched in Staging mode, waiting to be added to a running consensus, and thus, not electing itself as leader or doing anything like we were doing before. This additionally allows us to track when the peer has become a Voter, which only happens when it's caught up with the state, something that was wonky previously. * The raft configuration now includes an InitPeerset key, which allows to provide a peerset for new peers and which is ignored when staging==true. The whole Raft initialization code is way cleaner and stronger now. * Cluster peer bootsrapping is now an ipfs-cluster-service feature. The --bootstrap flag works as before (additionally allowing comma-separated-list of entries). What bootstrap does, is to initialize Raft with staging == true, and then call Join in the main cluster component. Only when the Raft peer transitions to Voter, consensus becomes ready, and cluster becomes Ready. This is cleaner, works better and is less complex than before (supporting both flags and config values). We also backup and clean the state whenever we are boostrapping, automatically * ipfs-cluster-service no longer runs the daemon. Starting cluster needs now "ipfs-cluster-service daemon". The daemon specific flags (bootstrap, alloc) are now flags for the daemon subcommand. Here we mimic ipfs ("ipfs" does not start the daemon but print help) and pave the path for merging both service and ctl in the future. While this brings some breaking changes, it significantly reduces the complexity of the configuration, the code and most importantly, the documentation. It should be easier now to explain the user what is the right way to launch a cluster peer, and more difficult to make mistakes. As a side effect, the PR also: * Fixes #381 - peers with dynamic addresses * Fixes #371 - peers should be Raft configuration option * Fixes #378 - waitForUpdates may return before state fully synced * Fixes #235 - config option shadowing (no cfg saves, no need to shadow) License: MIT Signed-off-by: Hector Sanjuan <code@hector.link>
2018-04-28 22:22:23 +00:00
diskInfCfg *disk.Config,
numpinInfCfg *numpin.Config,
) (ipfscluster.Informer, ipfscluster.PinAllocator) {
switch name {
case "disk", "disk-freespace":
informer, err := disk.NewInformer(diskInfCfg)
checkErr("creating informer", err)
return informer, descendalloc.NewAllocator()
case "disk-reposize":
informer, err := disk.NewInformer(diskInfCfg)
checkErr("creating informer", err)
return informer, ascendalloc.NewAllocator()
case "numpin", "pincount":
informer, err := numpin.NewInformer(numpinInfCfg)
checkErr("creating informer", err)
return informer, ascendalloc.NewAllocator()
default:
err := errors.New("unknown allocation strategy")
checkErr("", err)
return nil, nil
}
}
func setupMonitor(
name string,
h host.Host,
basicCfg *basic.Config,
pubsubCfg *pubsubmon.Config,
) ipfscluster.PeerMonitor {
switch name {
case "basic":
mon, err := basic.NewMonitor(basicCfg)
checkErr("creating monitor", err)
logger.Debug("basic monitor loaded")
return mon
case "pubsub":
mon, err := pubsubmon.New(h, pubsubCfg)
checkErr("creating monitor", err)
logger.Debug("pubsub monitor loaded")
return mon
default:
err := errors.New("unknown monitor type")
checkErr("", err)
return nil
}
}
func setupPinTracker(
name string,
h host.Host,
mapCfg *maptracker.Config,
statelessCfg *stateless.Config,
peerName string,
) ipfscluster.PinTracker {
switch name {
case "map":
ptrk := maptracker.NewMapPinTracker(mapCfg, h.ID(), peerName)
logger.Debug("map pintracker loaded")
return ptrk
case "stateless":
ptrk := stateless.New(statelessCfg, h.ID(), peerName)
logger.Debug("stateless pintracker loaded")
return ptrk
default:
err := errors.New("unknown pintracker type")
checkErr("", err)
return nil
}
}