my-rook-config/pkg/clusterd/servicesleader.go

/*
Copyright 2016 The Rook Authors. All rights reserved.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

	http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/
package clusterd

import (
	"time"

	etcd "github.com/coreos/etcd/client"
	"github.com/coreos/etcd/store"

	ctx "golang.org/x/net/context"

	"github.com/rook/rook/pkg/clusterd/inventory"
	"github.com/rook/rook/pkg/util"
)

var (
	detectUnhealthyNodeInterval = 5 * time.Second
)

type servicesLeader struct {
	leaseName     string
	context       *Context
	isLeader      bool
	parent        *ClusterMember
	watcherCancel ctx.CancelFunc
	refresher     *ClusterRefresher
}

func newServicesLeader(context *Context) *servicesLeader {
	l := &servicesLeader{context: context, refresher: NewClusterRefresher(), leaseName: LeaderElectionKey}
	l.refresher.leader = l
	return l
}

func (s *servicesLeader) OnLeadershipAcquired() error {
	s.onLeadershipAcquiredRefresh(true)
	return nil
}

func (s *servicesLeader) onLeadershipAcquiredRefresh(refresh bool) {
	s.isLeader = true

	// start watching for membership changes and handling any changes to it
	s.refresher.Start()
	s.startWatchingClusterChanges()
	s.startWatchingUnhealthyNodes()

	if refresh {
		// Tell the services to refresh the cluster whenever leadership is acquired
		s.refresher.TriggerRefresh()
	}
}

func (s *servicesLeader) OnLeadershipLost() error {
	s.isLeader = false
	s.stopWatchingClusterChanges()
	s.refresher.Stop()

	return nil
}

func (s *servicesLeader) GetLeaseName() string {
	return s.leaseName
}

func (s *servicesLeader) startWatchingUnhealthyNodes() {
	go func() {
		for {
			// look for unhealthy nodes at a regular interval
			<-time.After(detectUnhealthyNodeInterval)
			if !s.isLeader {
				// poor man's cancellation when leadership is lost
				break
			}

			err := s.discoverUnhealthyNodes()
			if err != nil {
				logger.Warningf("error while discovering unhealthy nodes: %+v", err)
			}
		}
	}()
}

func (s *servicesLeader) discoverUnhealthyNodes() error {
	// load the state of all the nodes in the cluster
	config, err := inventory.LoadDiscoveredNodes(s.context.EtcdClient)
	if err != nil {
		return err
	}

	// look for old heartbeats
	var unhealthyNodes []*UnhealthyNode
	for nodeID, node := range config.Nodes {
		age, unhealthy := IsNodeUnhealthy(node)
		if unhealthy {
			unhealthyNodes = append(unhealthyNodes, &UnhealthyNode{AgeSeconds: age, ID: nodeID})
		}
	}

	// if we found unhealthy nodes, raise an event
	if len(unhealthyNodes) > 0 {
		logger.Infof("Found %d unhealthy nodes", len(unhealthyNodes))
		s.refresher.triggerNodeUnhealthy(unhealthyNodes)
	}

	return nil
}

func handleNodeAdded(response *etcd.Response, refresher *ClusterRefresher) {
	if response.Action == store.Create {
		newNodeID := util.GetLeafKeyPath(response.Node.Key)
		logger.Noticef("new node discovered: %s", newNodeID)

		// trigger an orchestration to configure services on the new machine
		refresher.triggerNodeAdded(newNodeID)
	}
}

func (s *servicesLeader) startWatchingClusterChanges() {
	// create an etcd watcher object and initialize a cancellable context for it
	context, cancelFunc := ctx.WithCancel(ctx.Background())
	s.watcherCancel = cancelFunc

	// watch for changes in the discovered nodes etcd key
	nodeRefresh := &RefreshKey{Path: inventory.NodesConfigKey, Triggered: handleNodeAdded}
	go s.watchClusterChange(context, nodeRefresh)

	// watch for changes requested by the service managers
	for _, mgr := range s.context.Services {
		for _, refresh := range mgr.Leader.RefreshKeys() {
			go s.watchClusterChange(context, refresh)
		}
	}
}

func (s *servicesLeader) watchClusterChange(context ctx.Context, refreshKey *RefreshKey) {
	watcher := s.context.EtcdClient.Watcher(refreshKey.Path, &etcd.WatcherOptions{Recursive: true})
	for {
		logger.Tracef("watching cluster changes under %s", refreshKey.Path)
		resp, err := watcher.Next(context)
		if err != nil {
			if err == ctx.Canceled {
				logger.Debugf("%s change watching cancelled, bailing out...", refreshKey.Path)
				break
			} else {
				logger.Warningf(
					"%s change watcher Next returned error, sleeping %d sec before retry: %+v",
					refreshKey.Path,
					watchErrorRetrySeconds,
					err)
				<-time.After(time.Duration(watchErrorRetrySeconds) * time.Second)
				continue
			}
		}

		if resp != nil && resp.Node != nil {
			refreshKey.Triggered(resp, s.refresher)
		}
	}

}

func (s *servicesLeader) stopWatchingClusterChanges() {
	if s.watcherCancel != nil {
		logger.Infof("calling cancel function for cluster change watcher...")
		s.watcherCancel()
	}

	s.watcherCancel = nil
}