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

498 lines
14 KiB
Go
Raw Normal View History

// The ipfs-cluster-service application.
package main
import (
"bufio"
"fmt"
"io"
"os"
"os/user"
"path/filepath"
// _ "net/http/pprof"
ipfscluster "github.com/ipfs/ipfs-cluster"
"github.com/ipfs/ipfs-cluster/state/mapstate"
semver "github.com/blang/semver"
logging "github.com/ipfs/go-log"
cli "github.com/urfave/cli"
)
// ProgramName of this application
const programName = `ipfs-cluster-service`
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
// flag defaults
const (
defaultAllocation = "disk-freespace"
defaultMonitor = "pubsub"
defaultPinTracker = "map"
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
defaultLogLevel = "info"
)
// We store a commit id here
var commit string
// Description provides a short summary of the functionality of this tool
var Description = fmt.Sprintf(`
%s runs an IPFS Cluster node.
A node participates in the cluster consensus, follows a distributed log
of pinning and unpinning requests and manages pinning operations to a
configured IPFS daemon.
This node also provides an API for cluster management, an IPFS Proxy API which
forwards requests to IPFS and a number of components for internal communication
using LibP2P. This is a simplified view of the components:
+------------------+
| ipfs-cluster-ctl |
+---------+--------+
|
| HTTP(s)
ipfs-cluster-service | HTTP
+----------+--------+--v--+----------------------+ +-------------+
| RPC/Raft | Peer 1 | API | IPFS Connector/Proxy +------> IPFS daemon |
+----^-----+--------+-----+----------------------+ +-------------+
| libp2p
|
+----v-----+--------+-----+----------------------+ +-------------+
| RPC/Raft | Peer 2 | API | IPFS Connector/Proxy +------> IPFS daemon |
+----^-----+--------+-----+----------------------+ +-------------+
|
|
+----v-----+--------+-----+----------------------+ +-------------+
| RPC/Raft | Peer 3 | API | IPFS Connector/Proxy +------> IPFS daemon |
+----------+--------+-----+----------------------+ +-------------+
%s needs a valid configuration to run. This configuration is
independent from IPFS and includes its own LibP2P key-pair. It can be
initialized with "init" and its default location is
~/%s/%s.
For feedback, bug reports or any additional information, visit
https://github.com/ipfs/ipfs-cluster.
EXAMPLES
Initial configuration:
$ ipfs-cluster-service init
Launch a 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
$ ipfs-cluster-service daemon
Launch a peer and join existing 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
$ ipfs-cluster-service daemon --bootstrap /ip4/192.168.1.2/tcp/9096/ipfs/QmPSoSaPXpyunaBwHs1rZBKYSqRV4bLRk32VGYLuvdrypL
`,
programName,
programName,
DefaultPath,
DefaultConfigFile)
var logger = logging.Logger("service")
// Default location for the configurations and data
var (
// DefaultPath is initialized to $HOME/.ipfs-cluster
// and holds all the ipfs-cluster data
DefaultPath string
// The name of the configuration file inside DefaultPath
DefaultConfigFile = "service.json"
)
var (
configPath string
)
func init() {
// Set build information.
if build, err := semver.NewBuildVersion(commit); err == nil {
ipfscluster.Version.Build = []string{"git" + build}
}
// We try guessing user's home from the HOME variable. This
// allows HOME hacks for things like Snapcraft builds. HOME
// should be set in all UNIX by the OS. Alternatively, we fall back to
// usr.HomeDir (which should work on Windows etc.).
home := os.Getenv("HOME")
if home == "" {
usr, err := user.Current()
if err != nil {
panic(fmt.Sprintf("cannot get current user: %s", err))
}
home = usr.HomeDir
}
DefaultPath = filepath.Join(home, ".ipfs-cluster")
}
func out(m string, a ...interface{}) {
fmt.Fprintf(os.Stderr, m, a...)
}
func checkErr(doing string, err error, args ...interface{}) {
if err != nil {
if len(args) > 0 {
doing = fmt.Sprintf(doing, args)
}
out("error %s: %s\n", doing, err)
err = locker.tryUnlock()
if err != nil {
out("error releasing execution lock: %s\n", err)
}
os.Exit(1)
}
}
func main() {
// go func() {
// log.Println(http.ListenAndServe("localhost:6060", nil))
// }()
app := cli.NewApp()
app.Name = programName
app.Usage = "IPFS Cluster node"
app.Description = Description
//app.Copyright = "© Protocol Labs, Inc."
app.Version = ipfscluster.Version.String()
app.Flags = []cli.Flag{
cli.StringFlag{
Name: "config, c",
Value: DefaultPath,
Usage: "path to the configuration and data `FOLDER`",
EnvVar: "IPFS_CLUSTER_PATH",
},
cli.BoolFlag{
Name: "force, f",
Usage: "forcefully proceed with some actions. i.e. overwriting configuration",
},
cli.BoolFlag{
Name: "debug, d",
Usage: "enable full debug logging (very verbose)",
},
cli.StringFlag{
Name: "loglevel, l",
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
Value: defaultLogLevel,
Usage: "set the loglevel for cluster components only [critical, error, warning, info, debug]",
},
}
app.Commands = []cli.Command{
{
Name: "init",
Usage: "create a default configuration and exit",
Description: fmt.Sprintf(`
This command will initialize a new service.json configuration file
for %s.
By default, %s requires a cluster secret. This secret will be
automatically generated, but can be manually provided with --custom-secret
(in which case it will be prompted), or by setting the CLUSTER_SECRET
environment variable.
The private key for the libp2p node is randomly generated in all cases.
Note that the --force first-level-flag allows to overwrite an existing
configuration.
`, programName, programName),
ArgsUsage: " ",
Flags: []cli.Flag{
cli.BoolFlag{
Name: "custom-secret, s",
Usage: "prompt for the cluster secret",
},
},
Action: func(c *cli.Context) error {
userSecret, userSecretDefined := userProvidedSecret(c.Bool("custom-secret"))
cfgMgr, cfgs := makeConfigs()
defer cfgMgr.Shutdown() // wait for saves
Issue #162: Rework configuration format The following commit reimplements ipfs-cluster configuration under the following premises: * Each component is initialized with a configuration object defined by its module * Each component decides how the JSON representation of its configuration looks like * Each component parses and validates its own configuration * Each component exposes its own defaults * Component configurations are make the sections of a central JSON configuration file (which replaces the current JSON format) * Component configurations implement a common interface (config.ComponentConfig) with a set of common operations * The central configuration file is managed by a config.ConfigManager which: * Registers ComponentConfigs * Assigns the correspondent sections from the JSON file to each component and delegates the parsing * Delegates the JSON generation for each section * Can be notified when the configuration is updated and must be saved to disk The new service.json would then look as follows: ```json { "cluster": { "id": "QmTVW8NoRxC5wBhV7WtAYtRn7itipEESfozWN5KmXUQnk2", "private_key": "<...>", "secret": "00224102ae6aaf94f2606abf69a0e278251ecc1d64815b617ff19d6d2841f786", "peers": [], "bootstrap": [], "leave_on_shutdown": false, "listen_multiaddress": "/ip4/0.0.0.0/tcp/9096", "state_sync_interval": "1m0s", "ipfs_sync_interval": "2m10s", "replication_factor": -1, "monitor_ping_interval": "15s" }, "consensus": { "raft": { "heartbeat_timeout": "1s", "election_timeout": "1s", "commit_timeout": "50ms", "max_append_entries": 64, "trailing_logs": 10240, "snapshot_interval": "2m0s", "snapshot_threshold": 8192, "leader_lease_timeout": "500ms" } }, "api": { "restapi": { "listen_multiaddress": "/ip4/127.0.0.1/tcp/9094", "read_timeout": "30s", "read_header_timeout": "5s", "write_timeout": "1m0s", "idle_timeout": "2m0s" } }, "ipfs_connector": { "ipfshttp": { "proxy_listen_multiaddress": "/ip4/127.0.0.1/tcp/9095", "node_multiaddress": "/ip4/127.0.0.1/tcp/5001", "connect_swarms_delay": "7s", "proxy_read_timeout": "10m0s", "proxy_read_header_timeout": "5s", "proxy_write_timeout": "10m0s", "proxy_idle_timeout": "1m0s" } }, "monitor": { "monbasic": { "check_interval": "15s" } }, "informer": { "disk": { "metric_ttl": "30s", "metric_type": "freespace" }, "numpin": { "metric_ttl": "10s" } } } ``` This new format aims to be easily extensible per component. As such, it already surfaces quite a few new options which were hardcoded before. Additionally, since Go API have changed, some redundant methods have been removed and small refactoring has happened to take advantage of the new way. License: MIT Signed-off-by: Hector Sanjuan <hector@protocol.ai>
2017-10-11 18:23:03 +00:00
// Generate defaults for all registered components
err := cfgMgr.Default()
Issue #162: Rework configuration format The following commit reimplements ipfs-cluster configuration under the following premises: * Each component is initialized with a configuration object defined by its module * Each component decides how the JSON representation of its configuration looks like * Each component parses and validates its own configuration * Each component exposes its own defaults * Component configurations are make the sections of a central JSON configuration file (which replaces the current JSON format) * Component configurations implement a common interface (config.ComponentConfig) with a set of common operations * The central configuration file is managed by a config.ConfigManager which: * Registers ComponentConfigs * Assigns the correspondent sections from the JSON file to each component and delegates the parsing * Delegates the JSON generation for each section * Can be notified when the configuration is updated and must be saved to disk The new service.json would then look as follows: ```json { "cluster": { "id": "QmTVW8NoRxC5wBhV7WtAYtRn7itipEESfozWN5KmXUQnk2", "private_key": "<...>", "secret": "00224102ae6aaf94f2606abf69a0e278251ecc1d64815b617ff19d6d2841f786", "peers": [], "bootstrap": [], "leave_on_shutdown": false, "listen_multiaddress": "/ip4/0.0.0.0/tcp/9096", "state_sync_interval": "1m0s", "ipfs_sync_interval": "2m10s", "replication_factor": -1, "monitor_ping_interval": "15s" }, "consensus": { "raft": { "heartbeat_timeout": "1s", "election_timeout": "1s", "commit_timeout": "50ms", "max_append_entries": 64, "trailing_logs": 10240, "snapshot_interval": "2m0s", "snapshot_threshold": 8192, "leader_lease_timeout": "500ms" } }, "api": { "restapi": { "listen_multiaddress": "/ip4/127.0.0.1/tcp/9094", "read_timeout": "30s", "read_header_timeout": "5s", "write_timeout": "1m0s", "idle_timeout": "2m0s" } }, "ipfs_connector": { "ipfshttp": { "proxy_listen_multiaddress": "/ip4/127.0.0.1/tcp/9095", "node_multiaddress": "/ip4/127.0.0.1/tcp/5001", "connect_swarms_delay": "7s", "proxy_read_timeout": "10m0s", "proxy_read_header_timeout": "5s", "proxy_write_timeout": "10m0s", "proxy_idle_timeout": "1m0s" } }, "monitor": { "monbasic": { "check_interval": "15s" } }, "informer": { "disk": { "metric_ttl": "30s", "metric_type": "freespace" }, "numpin": { "metric_ttl": "10s" } } } ``` This new format aims to be easily extensible per component. As such, it already surfaces quite a few new options which were hardcoded before. Additionally, since Go API have changed, some redundant methods have been removed and small refactoring has happened to take advantage of the new way. License: MIT Signed-off-by: Hector Sanjuan <hector@protocol.ai>
2017-10-11 18:23:03 +00:00
checkErr("generating default configuration", err)
// Set user secret
if userSecretDefined {
cfgs.clusterCfg.Secret = userSecret
Issue #162: Rework configuration format The following commit reimplements ipfs-cluster configuration under the following premises: * Each component is initialized with a configuration object defined by its module * Each component decides how the JSON representation of its configuration looks like * Each component parses and validates its own configuration * Each component exposes its own defaults * Component configurations are make the sections of a central JSON configuration file (which replaces the current JSON format) * Component configurations implement a common interface (config.ComponentConfig) with a set of common operations * The central configuration file is managed by a config.ConfigManager which: * Registers ComponentConfigs * Assigns the correspondent sections from the JSON file to each component and delegates the parsing * Delegates the JSON generation for each section * Can be notified when the configuration is updated and must be saved to disk The new service.json would then look as follows: ```json { "cluster": { "id": "QmTVW8NoRxC5wBhV7WtAYtRn7itipEESfozWN5KmXUQnk2", "private_key": "<...>", "secret": "00224102ae6aaf94f2606abf69a0e278251ecc1d64815b617ff19d6d2841f786", "peers": [], "bootstrap": [], "leave_on_shutdown": false, "listen_multiaddress": "/ip4/0.0.0.0/tcp/9096", "state_sync_interval": "1m0s", "ipfs_sync_interval": "2m10s", "replication_factor": -1, "monitor_ping_interval": "15s" }, "consensus": { "raft": { "heartbeat_timeout": "1s", "election_timeout": "1s", "commit_timeout": "50ms", "max_append_entries": 64, "trailing_logs": 10240, "snapshot_interval": "2m0s", "snapshot_threshold": 8192, "leader_lease_timeout": "500ms" } }, "api": { "restapi": { "listen_multiaddress": "/ip4/127.0.0.1/tcp/9094", "read_timeout": "30s", "read_header_timeout": "5s", "write_timeout": "1m0s", "idle_timeout": "2m0s" } }, "ipfs_connector": { "ipfshttp": { "proxy_listen_multiaddress": "/ip4/127.0.0.1/tcp/9095", "node_multiaddress": "/ip4/127.0.0.1/tcp/5001", "connect_swarms_delay": "7s", "proxy_read_timeout": "10m0s", "proxy_read_header_timeout": "5s", "proxy_write_timeout": "10m0s", "proxy_idle_timeout": "1m0s" } }, "monitor": { "monbasic": { "check_interval": "15s" } }, "informer": { "disk": { "metric_ttl": "30s", "metric_type": "freespace" }, "numpin": { "metric_ttl": "10s" } } } ``` This new format aims to be easily extensible per component. As such, it already surfaces quite a few new options which were hardcoded before. Additionally, since Go API have changed, some redundant methods have been removed and small refactoring has happened to take advantage of the new way. License: MIT Signed-off-by: Hector Sanjuan <hector@protocol.ai>
2017-10-11 18:23:03 +00:00
}
// Save
saveConfig(cfgMgr, c.GlobalBool("force"))
return nil
},
},
{
Name: "daemon",
Usage: "run the IPFS Cluster peer (default)",
Flags: []cli.Flag{
cli.BoolFlag{
Name: "upgrade, u",
Usage: "run necessary state migrations before starting cluster service",
},
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
cli.StringSliceFlag{
Name: "bootstrap, j",
Usage: "join a cluster providing an existing peers multiaddress(es)",
},
cli.BoolFlag{
Name: "leave, x",
Usage: "remove peer from cluster on exit. Overrides \"leave_on_shutdown\"",
Hidden: true,
},
cli.StringFlag{
Name: "alloc, a",
Value: defaultAllocation,
Usage: "allocation strategy to use [disk-freespace,disk-reposize,numpin].",
},
cli.StringFlag{
Name: "monitor",
Value: defaultMonitor,
Hidden: true,
Usage: "peer monitor to use [basic,pubsub].",
},
cli.StringFlag{
Name: "pintracker",
Value: defaultPinTracker,
Hidden: true,
Usage: "pintracker to use [map,stateless].",
},
},
Action: daemon,
},
{
Name: "state",
Usage: "Manage ipfs-cluster-state",
Subcommands: []cli.Command{
{
Name: "version",
Usage: "display the shared state format version",
Action: func(c *cli.Context) error {
fmt.Printf("%d\n", mapstate.Version)
return nil
},
},
{
Name: "upgrade",
Usage: "upgrade the IPFS Cluster state to the current version",
Description: `
This command upgrades the internal state of the ipfs-cluster node
specified in the latest raft snapshot. The state format is migrated from the
version of the snapshot to the version supported by the current cluster version.
To successfully run an upgrade of an entire cluster, shut down each peer without
removal, upgrade state using this command, and restart every peer.
`,
Action: func(c *cli.Context) error {
err := locker.lock()
checkErr("acquiring execution lock", err)
defer locker.tryUnlock()
err = upgrade()
checkErr("upgrading state", err)
return nil
},
},
{
Name: "export",
Usage: "save the IPFS Cluster state to a json file",
Description: `
This command reads the current cluster state and saves it as json for
human readability and editing. Only state formats compatible with this
version of ipfs-cluster-service can be exported. By default this command
prints the state to stdout.
`,
Flags: []cli.Flag{
cli.StringFlag{
Name: "file, f",
Value: "",
Usage: "sets an output file for exported state",
},
},
Action: func(c *cli.Context) error {
err := locker.lock()
checkErr("acquiring execution lock", err)
defer locker.tryUnlock()
var w io.WriteCloser
outputPath := c.String("file")
if outputPath == "" {
// Output to stdout
w = os.Stdout
} else {
// Create the export file
w, err = os.Create(outputPath)
checkErr("creating output file", err)
}
defer w.Close()
err = export(w)
checkErr("exporting state", err)
return nil
},
},
{
Name: "import",
Usage: "load an IPFS Cluster state from an exported state file",
Description: `
This command reads in an exported state file storing the state as a persistent
snapshot to be loaded as the cluster state when the cluster peer is restarted.
If an argument is provided, cluster will treat it as the path of the file to
import. If no argument is provided cluster will read json from stdin
`,
Action: func(c *cli.Context) error {
err := locker.lock()
checkErr("acquiring execution lock", err)
defer locker.tryUnlock()
if !c.GlobalBool("force") {
if !yesNoPrompt("The peer's state will be replaced. Run with -h for details. Continue? [y/n]:") {
return nil
}
}
// Get the importing file path
importFile := c.Args().First()
var r io.ReadCloser
if importFile == "" {
r = os.Stdin
logger.Info("Reading from stdin, Ctrl-D to finish")
} else {
r, err = os.Open(importFile)
checkErr("reading import file", err)
}
defer r.Close()
err = stateImport(r)
checkErr("importing state", err)
logger.Info("the given state has been correctly imported to this peer. Make sure all peers have consistent states")
return nil
},
},
{
Name: "cleanup",
Usage: "cleanup persistent consensus state so cluster can start afresh",
Description: `
This command removes the persistent state that is loaded on startup to determine this peer's view of the
cluster state. While it removes the existing state from the load path, one invocation does not permanently remove
this state from disk. This command renames cluster's data folder to <data-folder-name>.old.0, and rotates other
deprecated data folders to <data-folder-name>.old.<n+1>, etc for some rotation factor before permanatly deleting
the mth data folder (m currently defaults to 5)
`,
Action: func(c *cli.Context) error {
err := locker.lock()
checkErr("acquiring execution lock", err)
defer locker.tryUnlock()
if !c.GlobalBool("force") {
if !yesNoPrompt("The peer's state will be removed from the load path. Existing pins may be lost. Continue? [y/n]:") {
return nil
}
}
cfgMgr, cfgs := makeConfigs()
err = cfgMgr.LoadJSONFromFile(configPath)
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("reading configuration", 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
err = cleanupState(cfgs.consensusCfg)
checkErr("Cleaning up consensus data", 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
logger.Warningf("the %s folder has been rotated. Next start will use an empty state", cfgs.consensusCfg.GetDataFolder())
return nil
},
},
},
},
{
Name: "version",
Usage: "Print the ipfs-cluster version",
Action: func(c *cli.Context) error {
fmt.Printf("%s\n", ipfscluster.Version)
return nil
},
},
}
app.Before = func(c *cli.Context) error {
absPath, err := filepath.Abs(c.String("config"))
if err != nil {
return err
}
configPath = filepath.Join(absPath, DefaultConfigFile)
setupLogLevel(c.String("loglevel"))
if c.Bool("debug") {
setupDebug()
}
locker = &lock{path: absPath}
return nil
}
app.Action = run
app.Run(os.Args)
}
// run daemon() by default, or error.
func run(c *cli.Context) 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
cli.ShowAppHelp(c)
os.Exit(1)
return nil
}
func setupLogLevel(lvl string) {
for f := range ipfscluster.LoggingFacilities {
ipfscluster.SetFacilityLogLevel(f, lvl)
}
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
ipfscluster.SetFacilityLogLevel("service", lvl)
}
func setupDebug() {
ipfscluster.SetFacilityLogLevel("*", "DEBUG")
}
func userProvidedSecret(enterSecret bool) ([]byte, bool) {
var secret string
if enterSecret {
secret = promptUser("Enter cluster secret (32-byte hex string): ")
} else if envSecret, envSecretDefined := os.LookupEnv("CLUSTER_SECRET"); envSecretDefined {
secret = envSecret
} else {
return nil, false
}
decodedSecret, err := ipfscluster.DecodeClusterSecret(secret)
checkErr("parsing user-provided secret", err)
return decodedSecret, true
}
func promptUser(msg string) string {
scanner := bufio.NewScanner(os.Stdin)
fmt.Print(msg)
scanner.Scan()
return scanner.Text()
}
2017-07-19 16:54:57 +00:00
// Lifted from go-ipfs/cmd/ipfs/daemon.go
func yesNoPrompt(prompt string) bool {
var s string
for i := 0; i < 3; i++ {
fmt.Printf("%s ", prompt)
fmt.Scanf("%s", &s)
switch s {
case "y", "Y":
return true
case "n", "N":
return false
case "":
return false
}
fmt.Println("Please press either 'y' or 'n'")
}
return false
}