// _ _ // __ _____ __ ___ ___ __ _| |_ ___ // \ \ /\ / / _ \/ _` \ \ / / |/ _` | __/ _ \ // \ V V / __/ (_| |\ V /| | (_| | || __/ // \_/\_/ \___|\__,_| \_/ |_|\__,_|\__\___| // // Copyright © 2016 - 2025 Weaviate B.V. All rights reserved. // // CONTACT: hello@weaviate.io // package bootstrap import ( "context" "fmt" "math/rand" "time" "github.com/getsentry/sentry-go" "github.com/sirupsen/logrus" cmd "github.com/weaviate/weaviate/cluster/proto/api" "github.com/weaviate/weaviate/cluster/resolver" entSentry "github.com/weaviate/weaviate/entities/sentry" ) // PeerJoiner is the interface we expect to be able to talk to the other peers to either Join or Notify them type PeerJoiner interface { Join(_ context.Context, leaderAddr string, _ *cmd.JoinPeerRequest) (*cmd.JoinPeerResponse, error) Notify(_ context.Context, leaderAddr string, _ *cmd.NotifyPeerRequest) (*cmd.NotifyPeerResponse, error) } // Bootstrapper is used to bootstrap this node by attempting to join it to a RAFT cluster. type Bootstrapper struct { peerJoiner PeerJoiner addrResolver resolver.ClusterStateReader isStoreReady func() bool localRaftAddr string localNodeID string voter bool retryPeriod time.Duration jitter time.Duration } // NewBootstrapper constructs a new bootsrapper func NewBootstrapper(peerJoiner PeerJoiner, raftID string, raftAddr string, voter bool, r resolver.ClusterStateReader, isStoreReady func() bool) *Bootstrapper { return &Bootstrapper{ peerJoiner: peerJoiner, addrResolver: r, retryPeriod: time.Second, jitter: time.Second, localNodeID: raftID, localRaftAddr: raftAddr, isStoreReady: isStoreReady, voter: voter, } } // Do iterates over a list of servers in an attempt to join this node to a cluster. func (b *Bootstrapper) Do(ctx context.Context, serverPortMap map[string]int, lg *logrus.Logger, stop chan struct{}) error { if entSentry.Enabled() { transaction := sentry.StartTransaction(ctx, "raft.bootstrap", sentry.WithOpName("init"), sentry.WithDescription("Attempt to bootstrap a raft cluster"), ) ctx = transaction.Context() defer transaction.Finish() } ticker := time.NewTicker(jitter(b.retryPeriod, b.jitter)) defer ticker.Stop() for { select { case <-stop: return nil case <-ctx.Done(): return ctx.Err() case <-ticker.C: if b.isStoreReady() { lg.WithField("action", "bootstrap").Info("node reporting ready, exiting bootstrap process") return nil } remoteNodes := ResolveRemoteNodes(b.addrResolver, serverPortMap) // We were not able to resolve any nodes to an address if len(remoteNodes) == 0 { lg.WithField("action", "bootstrap").WithField("join_list", serverPortMap).Warn("unable to resolve any node address to join") continue } // Always try to join an existing cluster first joiner := NewJoiner(b.peerJoiner, b.localNodeID, b.localRaftAddr, b.voter) if leader, err := joiner.Do(ctx, lg, remoteNodes); err != nil { lg.WithFields(logrus.Fields{ "action": "bootstrap", "servers": remoteNodes, "voter": b.voter, }).WithError(err).Warning("failed to join cluster") } else { lg.WithFields(logrus.Fields{ "action": "bootstrap", "leader": leader, }).Info("successfully joined cluster") return nil } // We are a voter, we resolve other peers but we're unable to join them. We're in the situation where we are // bootstrapping a new cluster and now we want to notify the other nodes. // Each node on notify will build a list of notified node. Once bootstrap expect is reached the nodes will // bootstrap together. if b.voter { // notify other servers about readiness of this node to be joined if err := b.notify(ctx, remoteNodes); err != nil { lg.WithFields(logrus.Fields{ "action": "bootstrap", "servers": remoteNodes, }).WithError(err).Error("failed to notify peers") continue } lg.WithFields(logrus.Fields{ "action": "bootstrap", "servers": remoteNodes, }).Info("notified peers this node is ready to join as voter") } } } } // notify attempts to notify all nodes in remoteNodes that this server is ready to bootstrap func (b *Bootstrapper) notify(ctx context.Context, remoteNodes map[string]string) (err error) { if entSentry.Enabled() { span := sentry.StartSpan(ctx, "raft.bootstrap.notify", sentry.WithOpName("notify"), sentry.WithDescription("Attempt to notify existing node(s) to join a cluster"), ) ctx = span.Context() span.SetData("servers", remoteNodes) defer span.Finish() } for _, addr := range remoteNodes { req := &cmd.NotifyPeerRequest{Id: b.localNodeID, Address: b.localRaftAddr} _, err = b.peerJoiner.Notify(ctx, addr, req) if err != nil { return err } } return } // ResolveRemoteNodes returns a list of remoteNodes addresses resolved using addrResolver. The nodes id used are // taken from serverPortMap keys and ports from the values func ResolveRemoteNodes(addrResolver resolver.ClusterStateReader, serverPortMap map[string]int) map[string]string { candidates := make(map[string]string, len(serverPortMap)) for name, raftPort := range serverPortMap { if addr := addrResolver.NodeAddress(name); addr != "" { candidates[name] = fmt.Sprintf("%s:%d", addr, raftPort) } } return candidates } // jitter introduce some jitter to a given duration d + [0, 1) * jit -> [d, d+jit] func jitter(d time.Duration, jit time.Duration) time.Duration { return d + time.Duration(float64(jit)*rand.Float64()) }