open-nomad/client/pluginmanager/csimanager/manager.go

241 lines
7.2 KiB
Go
Raw Normal View History

CSI Plugin Registration (#6555) This changeset implements the initial registration and fingerprinting of CSI Plugins as part of #5378. At a high level, it introduces the following: * A `csi_plugin` stanza as part of a Nomad task configuration, to allow a task to expose that it is a plugin. * A new task runner hook: `csi_plugin_supervisor`. This hook does two things. When the `csi_plugin` stanza is detected, it will automatically configure the plugin task to receive bidirectional mounts to the CSI intermediary directory. At runtime, it will then perform an initial heartbeat of the plugin and handle submitting it to the new `dynamicplugins.Registry` for further use by the client, and then run a lightweight heartbeat loop that will emit task events when health changes. * The `dynamicplugins.Registry` for handling plugins that run as Nomad tasks, in contrast to the existing catalog that requires `go-plugin` type plugins and to know the plugin configuration in advance. * The `csimanager` which fingerprints CSI plugins, in a similar way to `drivermanager` and `devicemanager`. It currently only fingerprints the NodeID from the plugin, and assumes that all plugins are monolithic. Missing features * We do not use the live updates of the `dynamicplugin` registry in the `csimanager` yet. * We do not deregister the plugins from the client when they shutdown yet, they just become indefinitely marked as unhealthy. This is deliberate until we figure out how we should manage deploying new versions of plugins/transitioning them.
2019-10-22 13:20:26 +00:00
package csimanager
import (
"context"
"fmt"
CSI Plugin Registration (#6555) This changeset implements the initial registration and fingerprinting of CSI Plugins as part of #5378. At a high level, it introduces the following: * A `csi_plugin` stanza as part of a Nomad task configuration, to allow a task to expose that it is a plugin. * A new task runner hook: `csi_plugin_supervisor`. This hook does two things. When the `csi_plugin` stanza is detected, it will automatically configure the plugin task to receive bidirectional mounts to the CSI intermediary directory. At runtime, it will then perform an initial heartbeat of the plugin and handle submitting it to the new `dynamicplugins.Registry` for further use by the client, and then run a lightweight heartbeat loop that will emit task events when health changes. * The `dynamicplugins.Registry` for handling plugins that run as Nomad tasks, in contrast to the existing catalog that requires `go-plugin` type plugins and to know the plugin configuration in advance. * The `csimanager` which fingerprints CSI plugins, in a similar way to `drivermanager` and `devicemanager`. It currently only fingerprints the NodeID from the plugin, and assumes that all plugins are monolithic. Missing features * We do not use the live updates of the `dynamicplugin` registry in the `csimanager` yet. * We do not deregister the plugins from the client when they shutdown yet, they just become indefinitely marked as unhealthy. This is deliberate until we figure out how we should manage deploying new versions of plugins/transitioning them.
2019-10-22 13:20:26 +00:00
"sync"
"time"
"github.com/hashicorp/go-hclog"
"github.com/hashicorp/nomad/client/dynamicplugins"
"github.com/hashicorp/nomad/client/pluginmanager"
"github.com/hashicorp/nomad/nomad/structs"
)
// defaultPluginResyncPeriod is the time interval used to do a full resync
// against the dynamicplugins, to account for missed updates.
const defaultPluginResyncPeriod = 30 * time.Second
// UpdateNodeCSIInfoFunc is the callback used to update the node from
// fingerprinting
type UpdateNodeCSIInfoFunc func(string, *structs.CSIInfo)
type TriggerNodeEvent func(*structs.NodeEvent)
CSI Plugin Registration (#6555) This changeset implements the initial registration and fingerprinting of CSI Plugins as part of #5378. At a high level, it introduces the following: * A `csi_plugin` stanza as part of a Nomad task configuration, to allow a task to expose that it is a plugin. * A new task runner hook: `csi_plugin_supervisor`. This hook does two things. When the `csi_plugin` stanza is detected, it will automatically configure the plugin task to receive bidirectional mounts to the CSI intermediary directory. At runtime, it will then perform an initial heartbeat of the plugin and handle submitting it to the new `dynamicplugins.Registry` for further use by the client, and then run a lightweight heartbeat loop that will emit task events when health changes. * The `dynamicplugins.Registry` for handling plugins that run as Nomad tasks, in contrast to the existing catalog that requires `go-plugin` type plugins and to know the plugin configuration in advance. * The `csimanager` which fingerprints CSI plugins, in a similar way to `drivermanager` and `devicemanager`. It currently only fingerprints the NodeID from the plugin, and assumes that all plugins are monolithic. Missing features * We do not use the live updates of the `dynamicplugin` registry in the `csimanager` yet. * We do not deregister the plugins from the client when they shutdown yet, they just become indefinitely marked as unhealthy. This is deliberate until we figure out how we should manage deploying new versions of plugins/transitioning them.
2019-10-22 13:20:26 +00:00
type Config struct {
Logger hclog.Logger
DynamicRegistry dynamicplugins.Registry
UpdateNodeCSIInfoFunc UpdateNodeCSIInfoFunc
PluginResyncPeriod time.Duration
TriggerNodeEvent TriggerNodeEvent
CSI Plugin Registration (#6555) This changeset implements the initial registration and fingerprinting of CSI Plugins as part of #5378. At a high level, it introduces the following: * A `csi_plugin` stanza as part of a Nomad task configuration, to allow a task to expose that it is a plugin. * A new task runner hook: `csi_plugin_supervisor`. This hook does two things. When the `csi_plugin` stanza is detected, it will automatically configure the plugin task to receive bidirectional mounts to the CSI intermediary directory. At runtime, it will then perform an initial heartbeat of the plugin and handle submitting it to the new `dynamicplugins.Registry` for further use by the client, and then run a lightweight heartbeat loop that will emit task events when health changes. * The `dynamicplugins.Registry` for handling plugins that run as Nomad tasks, in contrast to the existing catalog that requires `go-plugin` type plugins and to know the plugin configuration in advance. * The `csimanager` which fingerprints CSI plugins, in a similar way to `drivermanager` and `devicemanager`. It currently only fingerprints the NodeID from the plugin, and assumes that all plugins are monolithic. Missing features * We do not use the live updates of the `dynamicplugin` registry in the `csimanager` yet. * We do not deregister the plugins from the client when they shutdown yet, they just become indefinitely marked as unhealthy. This is deliberate until we figure out how we should manage deploying new versions of plugins/transitioning them.
2019-10-22 13:20:26 +00:00
}
// New returns a new PluginManager that will handle managing CSI plugins from
// the dynamicRegistry from the provided Config.
func New(config *Config) Manager {
CSI Plugin Registration (#6555) This changeset implements the initial registration and fingerprinting of CSI Plugins as part of #5378. At a high level, it introduces the following: * A `csi_plugin` stanza as part of a Nomad task configuration, to allow a task to expose that it is a plugin. * A new task runner hook: `csi_plugin_supervisor`. This hook does two things. When the `csi_plugin` stanza is detected, it will automatically configure the plugin task to receive bidirectional mounts to the CSI intermediary directory. At runtime, it will then perform an initial heartbeat of the plugin and handle submitting it to the new `dynamicplugins.Registry` for further use by the client, and then run a lightweight heartbeat loop that will emit task events when health changes. * The `dynamicplugins.Registry` for handling plugins that run as Nomad tasks, in contrast to the existing catalog that requires `go-plugin` type plugins and to know the plugin configuration in advance. * The `csimanager` which fingerprints CSI plugins, in a similar way to `drivermanager` and `devicemanager`. It currently only fingerprints the NodeID from the plugin, and assumes that all plugins are monolithic. Missing features * We do not use the live updates of the `dynamicplugin` registry in the `csimanager` yet. * We do not deregister the plugins from the client when they shutdown yet, they just become indefinitely marked as unhealthy. This is deliberate until we figure out how we should manage deploying new versions of plugins/transitioning them.
2019-10-22 13:20:26 +00:00
// Use a dedicated internal context for managing plugin shutdown.
ctx, cancelFn := context.WithCancel(context.Background())
if config.PluginResyncPeriod == 0 {
config.PluginResyncPeriod = defaultPluginResyncPeriod
}
return &csiManager{
logger: config.Logger,
eventer: config.TriggerNodeEvent,
CSI Plugin Registration (#6555) This changeset implements the initial registration and fingerprinting of CSI Plugins as part of #5378. At a high level, it introduces the following: * A `csi_plugin` stanza as part of a Nomad task configuration, to allow a task to expose that it is a plugin. * A new task runner hook: `csi_plugin_supervisor`. This hook does two things. When the `csi_plugin` stanza is detected, it will automatically configure the plugin task to receive bidirectional mounts to the CSI intermediary directory. At runtime, it will then perform an initial heartbeat of the plugin and handle submitting it to the new `dynamicplugins.Registry` for further use by the client, and then run a lightweight heartbeat loop that will emit task events when health changes. * The `dynamicplugins.Registry` for handling plugins that run as Nomad tasks, in contrast to the existing catalog that requires `go-plugin` type plugins and to know the plugin configuration in advance. * The `csimanager` which fingerprints CSI plugins, in a similar way to `drivermanager` and `devicemanager`. It currently only fingerprints the NodeID from the plugin, and assumes that all plugins are monolithic. Missing features * We do not use the live updates of the `dynamicplugin` registry in the `csimanager` yet. * We do not deregister the plugins from the client when they shutdown yet, they just become indefinitely marked as unhealthy. This is deliberate until we figure out how we should manage deploying new versions of plugins/transitioning them.
2019-10-22 13:20:26 +00:00
registry: config.DynamicRegistry,
instances: make(map[string]map[string]*instanceManager),
updateNodeCSIInfoFunc: config.UpdateNodeCSIInfoFunc,
pluginResyncPeriod: config.PluginResyncPeriod,
shutdownCtx: ctx,
shutdownCtxCancelFn: cancelFn,
shutdownCh: make(chan struct{}),
}
}
type csiManager struct {
// instances should only be accessed from the run() goroutine and the shutdown
// fn. It is a map of PluginType : [PluginName : *instanceManager]
CSI Plugin Registration (#6555) This changeset implements the initial registration and fingerprinting of CSI Plugins as part of #5378. At a high level, it introduces the following: * A `csi_plugin` stanza as part of a Nomad task configuration, to allow a task to expose that it is a plugin. * A new task runner hook: `csi_plugin_supervisor`. This hook does two things. When the `csi_plugin` stanza is detected, it will automatically configure the plugin task to receive bidirectional mounts to the CSI intermediary directory. At runtime, it will then perform an initial heartbeat of the plugin and handle submitting it to the new `dynamicplugins.Registry` for further use by the client, and then run a lightweight heartbeat loop that will emit task events when health changes. * The `dynamicplugins.Registry` for handling plugins that run as Nomad tasks, in contrast to the existing catalog that requires `go-plugin` type plugins and to know the plugin configuration in advance. * The `csimanager` which fingerprints CSI plugins, in a similar way to `drivermanager` and `devicemanager`. It currently only fingerprints the NodeID from the plugin, and assumes that all plugins are monolithic. Missing features * We do not use the live updates of the `dynamicplugin` registry in the `csimanager` yet. * We do not deregister the plugins from the client when they shutdown yet, they just become indefinitely marked as unhealthy. This is deliberate until we figure out how we should manage deploying new versions of plugins/transitioning them.
2019-10-22 13:20:26 +00:00
instances map[string]map[string]*instanceManager
registry dynamicplugins.Registry
logger hclog.Logger
eventer TriggerNodeEvent
CSI Plugin Registration (#6555) This changeset implements the initial registration and fingerprinting of CSI Plugins as part of #5378. At a high level, it introduces the following: * A `csi_plugin` stanza as part of a Nomad task configuration, to allow a task to expose that it is a plugin. * A new task runner hook: `csi_plugin_supervisor`. This hook does two things. When the `csi_plugin` stanza is detected, it will automatically configure the plugin task to receive bidirectional mounts to the CSI intermediary directory. At runtime, it will then perform an initial heartbeat of the plugin and handle submitting it to the new `dynamicplugins.Registry` for further use by the client, and then run a lightweight heartbeat loop that will emit task events when health changes. * The `dynamicplugins.Registry` for handling plugins that run as Nomad tasks, in contrast to the existing catalog that requires `go-plugin` type plugins and to know the plugin configuration in advance. * The `csimanager` which fingerprints CSI plugins, in a similar way to `drivermanager` and `devicemanager`. It currently only fingerprints the NodeID from the plugin, and assumes that all plugins are monolithic. Missing features * We do not use the live updates of the `dynamicplugin` registry in the `csimanager` yet. * We do not deregister the plugins from the client when they shutdown yet, they just become indefinitely marked as unhealthy. This is deliberate until we figure out how we should manage deploying new versions of plugins/transitioning them.
2019-10-22 13:20:26 +00:00
pluginResyncPeriod time.Duration
updateNodeCSIInfoFunc UpdateNodeCSIInfoFunc
shutdownCtx context.Context
shutdownCtxCancelFn context.CancelFunc
shutdownCh chan struct{}
}
func (c *csiManager) PluginManager() pluginmanager.PluginManager {
return c
}
func (c *csiManager) MounterForPlugin(ctx context.Context, pluginID string) (VolumeMounter, error) {
nodePlugins, hasAnyNodePlugins := c.instances["csi-node"]
if !hasAnyNodePlugins {
return nil, fmt.Errorf("no storage node plugins found")
}
mgr, hasPlugin := nodePlugins[pluginID]
if !hasPlugin {
return nil, fmt.Errorf("plugin %s for type csi-node not found", pluginID)
}
return mgr.VolumeMounter(ctx)
}
CSI Plugin Registration (#6555) This changeset implements the initial registration and fingerprinting of CSI Plugins as part of #5378. At a high level, it introduces the following: * A `csi_plugin` stanza as part of a Nomad task configuration, to allow a task to expose that it is a plugin. * A new task runner hook: `csi_plugin_supervisor`. This hook does two things. When the `csi_plugin` stanza is detected, it will automatically configure the plugin task to receive bidirectional mounts to the CSI intermediary directory. At runtime, it will then perform an initial heartbeat of the plugin and handle submitting it to the new `dynamicplugins.Registry` for further use by the client, and then run a lightweight heartbeat loop that will emit task events when health changes. * The `dynamicplugins.Registry` for handling plugins that run as Nomad tasks, in contrast to the existing catalog that requires `go-plugin` type plugins and to know the plugin configuration in advance. * The `csimanager` which fingerprints CSI plugins, in a similar way to `drivermanager` and `devicemanager`. It currently only fingerprints the NodeID from the plugin, and assumes that all plugins are monolithic. Missing features * We do not use the live updates of the `dynamicplugin` registry in the `csimanager` yet. * We do not deregister the plugins from the client when they shutdown yet, they just become indefinitely marked as unhealthy. This is deliberate until we figure out how we should manage deploying new versions of plugins/transitioning them.
2019-10-22 13:20:26 +00:00
// Run starts a plugin manager and should return early
func (c *csiManager) Run() {
go c.runLoop()
}
func (c *csiManager) runLoop() {
timer := time.NewTimer(0) // ensure we sync immediately in first pass
controllerUpdates := c.registry.PluginsUpdatedCh(c.shutdownCtx, "csi-controller")
nodeUpdates := c.registry.PluginsUpdatedCh(c.shutdownCtx, "csi-node")
CSI Plugin Registration (#6555) This changeset implements the initial registration and fingerprinting of CSI Plugins as part of #5378. At a high level, it introduces the following: * A `csi_plugin` stanza as part of a Nomad task configuration, to allow a task to expose that it is a plugin. * A new task runner hook: `csi_plugin_supervisor`. This hook does two things. When the `csi_plugin` stanza is detected, it will automatically configure the plugin task to receive bidirectional mounts to the CSI intermediary directory. At runtime, it will then perform an initial heartbeat of the plugin and handle submitting it to the new `dynamicplugins.Registry` for further use by the client, and then run a lightweight heartbeat loop that will emit task events when health changes. * The `dynamicplugins.Registry` for handling plugins that run as Nomad tasks, in contrast to the existing catalog that requires `go-plugin` type plugins and to know the plugin configuration in advance. * The `csimanager` which fingerprints CSI plugins, in a similar way to `drivermanager` and `devicemanager`. It currently only fingerprints the NodeID from the plugin, and assumes that all plugins are monolithic. Missing features * We do not use the live updates of the `dynamicplugin` registry in the `csimanager` yet. * We do not deregister the plugins from the client when they shutdown yet, they just become indefinitely marked as unhealthy. This is deliberate until we figure out how we should manage deploying new versions of plugins/transitioning them.
2019-10-22 13:20:26 +00:00
for {
select {
case <-timer.C:
c.resyncPluginsFromRegistry("csi-controller")
c.resyncPluginsFromRegistry("csi-node")
timer.Reset(c.pluginResyncPeriod)
case event := <-controllerUpdates:
c.handlePluginEvent(event)
case event := <-nodeUpdates:
c.handlePluginEvent(event)
case <-c.shutdownCtx.Done():
close(c.shutdownCh)
return
CSI Plugin Registration (#6555) This changeset implements the initial registration and fingerprinting of CSI Plugins as part of #5378. At a high level, it introduces the following: * A `csi_plugin` stanza as part of a Nomad task configuration, to allow a task to expose that it is a plugin. * A new task runner hook: `csi_plugin_supervisor`. This hook does two things. When the `csi_plugin` stanza is detected, it will automatically configure the plugin task to receive bidirectional mounts to the CSI intermediary directory. At runtime, it will then perform an initial heartbeat of the plugin and handle submitting it to the new `dynamicplugins.Registry` for further use by the client, and then run a lightweight heartbeat loop that will emit task events when health changes. * The `dynamicplugins.Registry` for handling plugins that run as Nomad tasks, in contrast to the existing catalog that requires `go-plugin` type plugins and to know the plugin configuration in advance. * The `csimanager` which fingerprints CSI plugins, in a similar way to `drivermanager` and `devicemanager`. It currently only fingerprints the NodeID from the plugin, and assumes that all plugins are monolithic. Missing features * We do not use the live updates of the `dynamicplugin` registry in the `csimanager` yet. * We do not deregister the plugins from the client when they shutdown yet, they just become indefinitely marked as unhealthy. This is deliberate until we figure out how we should manage deploying new versions of plugins/transitioning them.
2019-10-22 13:20:26 +00:00
}
}
}
// resyncPluginsFromRegistry does a full sync of the running instance
// managers against those in the registry. we primarily will use update
// events from the registry.
CSI Plugin Registration (#6555) This changeset implements the initial registration and fingerprinting of CSI Plugins as part of #5378. At a high level, it introduces the following: * A `csi_plugin` stanza as part of a Nomad task configuration, to allow a task to expose that it is a plugin. * A new task runner hook: `csi_plugin_supervisor`. This hook does two things. When the `csi_plugin` stanza is detected, it will automatically configure the plugin task to receive bidirectional mounts to the CSI intermediary directory. At runtime, it will then perform an initial heartbeat of the plugin and handle submitting it to the new `dynamicplugins.Registry` for further use by the client, and then run a lightweight heartbeat loop that will emit task events when health changes. * The `dynamicplugins.Registry` for handling plugins that run as Nomad tasks, in contrast to the existing catalog that requires `go-plugin` type plugins and to know the plugin configuration in advance. * The `csimanager` which fingerprints CSI plugins, in a similar way to `drivermanager` and `devicemanager`. It currently only fingerprints the NodeID from the plugin, and assumes that all plugins are monolithic. Missing features * We do not use the live updates of the `dynamicplugin` registry in the `csimanager` yet. * We do not deregister the plugins from the client when they shutdown yet, they just become indefinitely marked as unhealthy. This is deliberate until we figure out how we should manage deploying new versions of plugins/transitioning them.
2019-10-22 13:20:26 +00:00
func (c *csiManager) resyncPluginsFromRegistry(ptype string) {
plugins := c.registry.ListPlugins(ptype)
seen := make(map[string]struct{}, len(plugins))
// For every plugin in the registry, ensure that we have an existing plugin
// running. Also build the map of valid plugin names.
// Note: monolith plugins that run as both controllers and nodes get a
// separate instance manager for both modes.
CSI Plugin Registration (#6555) This changeset implements the initial registration and fingerprinting of CSI Plugins as part of #5378. At a high level, it introduces the following: * A `csi_plugin` stanza as part of a Nomad task configuration, to allow a task to expose that it is a plugin. * A new task runner hook: `csi_plugin_supervisor`. This hook does two things. When the `csi_plugin` stanza is detected, it will automatically configure the plugin task to receive bidirectional mounts to the CSI intermediary directory. At runtime, it will then perform an initial heartbeat of the plugin and handle submitting it to the new `dynamicplugins.Registry` for further use by the client, and then run a lightweight heartbeat loop that will emit task events when health changes. * The `dynamicplugins.Registry` for handling plugins that run as Nomad tasks, in contrast to the existing catalog that requires `go-plugin` type plugins and to know the plugin configuration in advance. * The `csimanager` which fingerprints CSI plugins, in a similar way to `drivermanager` and `devicemanager`. It currently only fingerprints the NodeID from the plugin, and assumes that all plugins are monolithic. Missing features * We do not use the live updates of the `dynamicplugin` registry in the `csimanager` yet. * We do not deregister the plugins from the client when they shutdown yet, they just become indefinitely marked as unhealthy. This is deliberate until we figure out how we should manage deploying new versions of plugins/transitioning them.
2019-10-22 13:20:26 +00:00
for _, plugin := range plugins {
seen[plugin.Name] = struct{}{}
c.ensureInstance(plugin)
CSI Plugin Registration (#6555) This changeset implements the initial registration and fingerprinting of CSI Plugins as part of #5378. At a high level, it introduces the following: * A `csi_plugin` stanza as part of a Nomad task configuration, to allow a task to expose that it is a plugin. * A new task runner hook: `csi_plugin_supervisor`. This hook does two things. When the `csi_plugin` stanza is detected, it will automatically configure the plugin task to receive bidirectional mounts to the CSI intermediary directory. At runtime, it will then perform an initial heartbeat of the plugin and handle submitting it to the new `dynamicplugins.Registry` for further use by the client, and then run a lightweight heartbeat loop that will emit task events when health changes. * The `dynamicplugins.Registry` for handling plugins that run as Nomad tasks, in contrast to the existing catalog that requires `go-plugin` type plugins and to know the plugin configuration in advance. * The `csimanager` which fingerprints CSI plugins, in a similar way to `drivermanager` and `devicemanager`. It currently only fingerprints the NodeID from the plugin, and assumes that all plugins are monolithic. Missing features * We do not use the live updates of the `dynamicplugin` registry in the `csimanager` yet. * We do not deregister the plugins from the client when they shutdown yet, they just become indefinitely marked as unhealthy. This is deliberate until we figure out how we should manage deploying new versions of plugins/transitioning them.
2019-10-22 13:20:26 +00:00
}
// For every instance manager, if we did not find it during the plugin
// iterator, shut it down and remove it from the table.
instances := c.instancesForType(ptype)
for name, mgr := range instances {
CSI Plugin Registration (#6555) This changeset implements the initial registration and fingerprinting of CSI Plugins as part of #5378. At a high level, it introduces the following: * A `csi_plugin` stanza as part of a Nomad task configuration, to allow a task to expose that it is a plugin. * A new task runner hook: `csi_plugin_supervisor`. This hook does two things. When the `csi_plugin` stanza is detected, it will automatically configure the plugin task to receive bidirectional mounts to the CSI intermediary directory. At runtime, it will then perform an initial heartbeat of the plugin and handle submitting it to the new `dynamicplugins.Registry` for further use by the client, and then run a lightweight heartbeat loop that will emit task events when health changes. * The `dynamicplugins.Registry` for handling plugins that run as Nomad tasks, in contrast to the existing catalog that requires `go-plugin` type plugins and to know the plugin configuration in advance. * The `csimanager` which fingerprints CSI plugins, in a similar way to `drivermanager` and `devicemanager`. It currently only fingerprints the NodeID from the plugin, and assumes that all plugins are monolithic. Missing features * We do not use the live updates of the `dynamicplugin` registry in the `csimanager` yet. * We do not deregister the plugins from the client when they shutdown yet, they just become indefinitely marked as unhealthy. This is deliberate until we figure out how we should manage deploying new versions of plugins/transitioning them.
2019-10-22 13:20:26 +00:00
if _, ok := seen[name]; !ok {
c.ensureNoInstance(mgr.info)
CSI Plugin Registration (#6555) This changeset implements the initial registration and fingerprinting of CSI Plugins as part of #5378. At a high level, it introduces the following: * A `csi_plugin` stanza as part of a Nomad task configuration, to allow a task to expose that it is a plugin. * A new task runner hook: `csi_plugin_supervisor`. This hook does two things. When the `csi_plugin` stanza is detected, it will automatically configure the plugin task to receive bidirectional mounts to the CSI intermediary directory. At runtime, it will then perform an initial heartbeat of the plugin and handle submitting it to the new `dynamicplugins.Registry` for further use by the client, and then run a lightweight heartbeat loop that will emit task events when health changes. * The `dynamicplugins.Registry` for handling plugins that run as Nomad tasks, in contrast to the existing catalog that requires `go-plugin` type plugins and to know the plugin configuration in advance. * The `csimanager` which fingerprints CSI plugins, in a similar way to `drivermanager` and `devicemanager`. It currently only fingerprints the NodeID from the plugin, and assumes that all plugins are monolithic. Missing features * We do not use the live updates of the `dynamicplugin` registry in the `csimanager` yet. * We do not deregister the plugins from the client when they shutdown yet, they just become indefinitely marked as unhealthy. This is deliberate until we figure out how we should manage deploying new versions of plugins/transitioning them.
2019-10-22 13:20:26 +00:00
}
}
}
// handlePluginEvent syncs a single event against the plugin registry
func (c *csiManager) handlePluginEvent(event *dynamicplugins.PluginUpdateEvent) {
if event == nil {
return
}
c.logger.Trace("dynamic plugin event",
"event", event.EventType,
"plugin_id", event.Info.Name,
"plugin_alloc_id", event.Info.AllocID)
switch event.EventType {
case dynamicplugins.EventTypeRegistered:
c.ensureInstance(event.Info)
case dynamicplugins.EventTypeDeregistered:
c.ensureNoInstance(event.Info)
default:
c.logger.Error("received unknown dynamic plugin event type",
"type", event.EventType)
}
}
// Ensure we have an instance manager for the plugin and add it to
// the CSI manager's tracking table for that plugin type.
func (c *csiManager) ensureInstance(plugin *dynamicplugins.PluginInfo) {
name := plugin.Name
ptype := plugin.Type
instances := c.instancesForType(ptype)
mgr, ok := instances[name]
if !ok {
c.logger.Debug("detected new CSI plugin", "name", name, "type", ptype, "alloc", plugin.AllocID)
mgr := newInstanceManager(c.logger, c.eventer, c.updateNodeCSIInfoFunc, plugin)
instances[name] = mgr
mgr.run()
} else if mgr.allocID != plugin.AllocID {
mgr.shutdown()
c.logger.Debug("detected update for CSI plugin", "name", name, "type", ptype, "alloc", plugin.AllocID)
mgr := newInstanceManager(c.logger, c.eventer, c.updateNodeCSIInfoFunc, plugin)
instances[name] = mgr
mgr.run()
}
}
// Shut down the instance manager for a plugin and remove it from
// the CSI manager's tracking table for that plugin type.
func (c *csiManager) ensureNoInstance(plugin *dynamicplugins.PluginInfo) {
name := plugin.Name
ptype := plugin.Type
instances := c.instancesForType(ptype)
if mgr, ok := instances[name]; ok {
if mgr.allocID == plugin.AllocID {
c.logger.Debug("shutting down CSI plugin", "name", name, "type", ptype, "alloc", plugin.AllocID)
mgr.shutdown()
delete(instances, name)
}
}
}
// Get the instance managers table for a specific plugin type,
// ensuring it's been initialized if it doesn't exist.
func (c *csiManager) instancesForType(ptype string) map[string]*instanceManager {
pluginMap, ok := c.instances[ptype]
if !ok {
pluginMap = make(map[string]*instanceManager)
c.instances[ptype] = pluginMap
}
return pluginMap
}
CSI Plugin Registration (#6555) This changeset implements the initial registration and fingerprinting of CSI Plugins as part of #5378. At a high level, it introduces the following: * A `csi_plugin` stanza as part of a Nomad task configuration, to allow a task to expose that it is a plugin. * A new task runner hook: `csi_plugin_supervisor`. This hook does two things. When the `csi_plugin` stanza is detected, it will automatically configure the plugin task to receive bidirectional mounts to the CSI intermediary directory. At runtime, it will then perform an initial heartbeat of the plugin and handle submitting it to the new `dynamicplugins.Registry` for further use by the client, and then run a lightweight heartbeat loop that will emit task events when health changes. * The `dynamicplugins.Registry` for handling plugins that run as Nomad tasks, in contrast to the existing catalog that requires `go-plugin` type plugins and to know the plugin configuration in advance. * The `csimanager` which fingerprints CSI plugins, in a similar way to `drivermanager` and `devicemanager`. It currently only fingerprints the NodeID from the plugin, and assumes that all plugins are monolithic. Missing features * We do not use the live updates of the `dynamicplugin` registry in the `csimanager` yet. * We do not deregister the plugins from the client when they shutdown yet, they just become indefinitely marked as unhealthy. This is deliberate until we figure out how we should manage deploying new versions of plugins/transitioning them.
2019-10-22 13:20:26 +00:00
// Shutdown should gracefully shutdown all plugins managed by the manager.
// It must block until shutdown is complete
func (c *csiManager) Shutdown() {
// Shut down the run loop
c.shutdownCtxCancelFn()
// Wait for plugin manager shutdown to complete so that we
// don't try to shutdown instance managers while runLoop is
// doing a resync
CSI Plugin Registration (#6555) This changeset implements the initial registration and fingerprinting of CSI Plugins as part of #5378. At a high level, it introduces the following: * A `csi_plugin` stanza as part of a Nomad task configuration, to allow a task to expose that it is a plugin. * A new task runner hook: `csi_plugin_supervisor`. This hook does two things. When the `csi_plugin` stanza is detected, it will automatically configure the plugin task to receive bidirectional mounts to the CSI intermediary directory. At runtime, it will then perform an initial heartbeat of the plugin and handle submitting it to the new `dynamicplugins.Registry` for further use by the client, and then run a lightweight heartbeat loop that will emit task events when health changes. * The `dynamicplugins.Registry` for handling plugins that run as Nomad tasks, in contrast to the existing catalog that requires `go-plugin` type plugins and to know the plugin configuration in advance. * The `csimanager` which fingerprints CSI plugins, in a similar way to `drivermanager` and `devicemanager`. It currently only fingerprints the NodeID from the plugin, and assumes that all plugins are monolithic. Missing features * We do not use the live updates of the `dynamicplugin` registry in the `csimanager` yet. * We do not deregister the plugins from the client when they shutdown yet, they just become indefinitely marked as unhealthy. This is deliberate until we figure out how we should manage deploying new versions of plugins/transitioning them.
2019-10-22 13:20:26 +00:00
<-c.shutdownCh
// Shutdown all the instance managers in parallel
var wg sync.WaitGroup
for _, pluginMap := range c.instances {
for _, mgr := range pluginMap {
wg.Add(1)
go func(mgr *instanceManager) {
mgr.shutdown()
wg.Done()
}(mgr)
}
}
wg.Wait()
}
// PluginType is the type of plugin which the manager manages
func (c *csiManager) PluginType() string {
return "csi"
}