diff options
Diffstat (limited to '')
-rw-r--r-- | pkg/icingadb/ha.go | 446 |
1 files changed, 446 insertions, 0 deletions
diff --git a/pkg/icingadb/ha.go b/pkg/icingadb/ha.go new file mode 100644 index 0000000..0a81bb5 --- /dev/null +++ b/pkg/icingadb/ha.go @@ -0,0 +1,446 @@ +package icingadb + +import ( + "bytes" + "context" + "database/sql" + "encoding/hex" + "github.com/google/uuid" + "github.com/icinga/icingadb/internal" + "github.com/icinga/icingadb/pkg/backoff" + "github.com/icinga/icingadb/pkg/com" + v1 "github.com/icinga/icingadb/pkg/icingadb/v1" + "github.com/icinga/icingadb/pkg/icingaredis" + icingaredisv1 "github.com/icinga/icingadb/pkg/icingaredis/v1" + "github.com/icinga/icingadb/pkg/logging" + "github.com/icinga/icingadb/pkg/retry" + "github.com/icinga/icingadb/pkg/types" + "github.com/icinga/icingadb/pkg/utils" + "github.com/pkg/errors" + "go.uber.org/zap" + "sync" + "time" +) + +var timeout = 60 * time.Second + +type haState struct { + responsibleTsMilli int64 + responsible bool + otherResponsible bool +} + +// HA provides high availability and indicates whether a Takeover or Handover must be made. +type HA struct { + state com.Atomic[haState] + ctx context.Context + cancelCtx context.CancelFunc + instanceId types.Binary + db *DB + environmentMu sync.Mutex + environment *v1.Environment + heartbeat *icingaredis.Heartbeat + logger *logging.Logger + responsible bool + handover chan struct{} + takeover chan struct{} + done chan struct{} + errOnce sync.Once + errMu sync.Mutex + err error +} + +// NewHA returns a new HA and starts the controller loop. +func NewHA(ctx context.Context, db *DB, heartbeat *icingaredis.Heartbeat, logger *logging.Logger) *HA { + ctx, cancelCtx := context.WithCancel(ctx) + + instanceId := uuid.New() + + ha := &HA{ + ctx: ctx, + cancelCtx: cancelCtx, + instanceId: instanceId[:], + db: db, + heartbeat: heartbeat, + logger: logger, + handover: make(chan struct{}), + takeover: make(chan struct{}), + done: make(chan struct{}), + } + + go ha.controller() + + return ha +} + +// Close shuts h down. +func (h *HA) Close(ctx context.Context) error { + // Cancel ctx. + h.cancelCtx() + // Wait until the controller loop ended. + <-h.Done() + // Remove our instance from the database. + h.removeInstance(ctx) + // And return an error, if any. + return h.Err() +} + +// Done returns a channel that's closed when the HA controller loop ended. +func (h *HA) Done() <-chan struct{} { + return h.done +} + +// Environment returns the current environment. +func (h *HA) Environment() *v1.Environment { + h.environmentMu.Lock() + defer h.environmentMu.Unlock() + + return h.environment +} + +// Err returns an error if Done has been closed and there is an error. Otherwise returns nil. +func (h *HA) Err() error { + h.errMu.Lock() + defer h.errMu.Unlock() + + return h.err +} + +// Handover returns a channel with which handovers are signaled. +func (h *HA) Handover() chan struct{} { + return h.handover +} + +// Takeover returns a channel with which takeovers are signaled. +func (h *HA) Takeover() chan struct{} { + return h.takeover +} + +// State returns the status quo. +func (h *HA) State() (responsibleTsMilli int64, responsible, otherResponsible bool) { + state, _ := h.state.Load() + return state.responsibleTsMilli, state.responsible, state.otherResponsible +} + +func (h *HA) abort(err error) { + h.errOnce.Do(func() { + h.errMu.Lock() + h.err = errors.Wrap(err, "HA aborted") + h.errMu.Unlock() + + h.cancelCtx() + }) +} + +// controller loop. +func (h *HA) controller() { + defer close(h.done) + + h.logger.Debugw("Starting HA", zap.String("instance_id", hex.EncodeToString(h.instanceId))) + + oldInstancesRemoved := false + + logTicker := time.NewTicker(time.Second * 60) + defer logTicker.Stop() + shouldLog := true + + for { + select { + case m := <-h.heartbeat.Events(): + if m != nil { + now := time.Now() + t, err := m.Stats().Time() + if err != nil { + h.abort(err) + } + tt := t.Time() + if tt.After(now.Add(1 * time.Second)) { + h.logger.Debugw("Received heartbeat from the future", zap.Time("time", tt)) + } + if tt.Before(now.Add(-1 * timeout)) { + h.logger.Errorw("Received heartbeat from the past", zap.Time("time", tt)) + h.signalHandover() + h.realizeLostHeartbeat() + continue + } + s, err := m.Stats().IcingaStatus() + if err != nil { + h.abort(err) + } + + envId, err := m.EnvironmentID() + if err != nil { + h.abort(err) + } + + if h.environment == nil || !bytes.Equal(h.environment.Id, envId) { + if h.environment != nil { + h.logger.Fatalw("Environment changed unexpectedly", + zap.String("current", h.environment.Id.String()), + zap.String("new", envId.String())) + } + + h.environmentMu.Lock() + h.environment = &v1.Environment{ + EntityWithoutChecksum: v1.EntityWithoutChecksum{IdMeta: v1.IdMeta{ + Id: envId, + }}, + Name: types.String{ + NullString: sql.NullString{ + String: envId.String(), + Valid: true, + }, + }, + } + h.environmentMu.Unlock() + } + + select { + case <-logTicker.C: + shouldLog = true + default: + } + + var realizeCtx context.Context + var cancelRealizeCtx context.CancelFunc + if h.responsible { + realizeCtx, cancelRealizeCtx = context.WithDeadline(h.ctx, m.ExpiryTime()) + } else { + realizeCtx, cancelRealizeCtx = context.WithCancel(h.ctx) + } + err = h.realize(realizeCtx, s, t, envId, shouldLog) + cancelRealizeCtx() + if errors.Is(err, context.DeadlineExceeded) { + h.signalHandover() + continue + } + if err != nil { + h.abort(err) + } + + if !oldInstancesRemoved { + go h.removeOldInstances(s, envId) + oldInstancesRemoved = true + } + + shouldLog = false + } else { + h.logger.Error("Lost heartbeat") + h.signalHandover() + h.realizeLostHeartbeat() + } + case <-h.heartbeat.Done(): + if err := h.heartbeat.Err(); err != nil { + h.abort(err) + } + case <-h.ctx.Done(): + return + } + } +} + +func (h *HA) realize(ctx context.Context, s *icingaredisv1.IcingaStatus, t *types.UnixMilli, envId types.Binary, shouldLog bool) error { + var takeover, otherResponsible bool + + err := retry.WithBackoff( + ctx, + func(ctx context.Context) error { + takeover = false + otherResponsible = false + + tx, errBegin := h.db.BeginTxx(ctx, &sql.TxOptions{Isolation: sql.LevelSerializable}) + if errBegin != nil { + return errors.Wrap(errBegin, "can't start transaction") + } + + query := h.db.Rebind("SELECT id, heartbeat FROM icingadb_instance " + + "WHERE environment_id = ? AND responsible = ? AND id <> ? AND heartbeat > ?") + + instance := &v1.IcingadbInstance{} + + errQuery := tx.QueryRowxContext( + ctx, query, envId, "y", h.instanceId, time.Now().Add(-1*timeout).UnixMilli(), + ).StructScan(instance) + switch errQuery { + case nil: + otherResponsible = true + if shouldLog { + h.logger.Infow("Another instance is active", + zap.String("instance_id", instance.Id.String()), + zap.String("environment", envId.String()), + "heartbeat", instance.Heartbeat, + zap.Duration("heartbeat_age", time.Since(instance.Heartbeat.Time()))) + } + case sql.ErrNoRows: + takeover = true + default: + return internal.CantPerformQuery(errQuery, query) + } + + i := v1.IcingadbInstance{ + EntityWithoutChecksum: v1.EntityWithoutChecksum{ + IdMeta: v1.IdMeta{ + Id: h.instanceId, + }, + }, + EnvironmentMeta: v1.EnvironmentMeta{ + EnvironmentId: envId, + }, + Heartbeat: *t, + Responsible: types.Bool{Bool: takeover || h.responsible, Valid: true}, + EndpointId: s.EndpointId, + Icinga2Version: s.Version, + Icinga2StartTime: s.ProgramStart, + Icinga2NotificationsEnabled: s.NotificationsEnabled, + Icinga2ActiveServiceChecksEnabled: s.ActiveServiceChecksEnabled, + Icinga2ActiveHostChecksEnabled: s.ActiveHostChecksEnabled, + Icinga2EventHandlersEnabled: s.EventHandlersEnabled, + Icinga2FlapDetectionEnabled: s.FlapDetectionEnabled, + Icinga2PerformanceDataEnabled: s.PerformanceDataEnabled, + } + + stmt, _ := h.db.BuildUpsertStmt(i) + if _, err := tx.NamedExecContext(ctx, stmt, i); err != nil { + return internal.CantPerformQuery(err, stmt) + } + + if takeover { + stmt := h.db.Rebind("UPDATE icingadb_instance SET responsible = ? WHERE environment_id = ? AND id <> ?") + _, err := tx.ExecContext(ctx, stmt, "n", envId, h.instanceId) + + if err != nil { + return internal.CantPerformQuery(err, stmt) + } + } + + if err := tx.Commit(); err != nil { + return errors.Wrap(err, "can't commit transaction") + } + + return nil + }, + IsRetryable, + backoff.NewExponentialWithJitter(time.Millisecond*256, time.Second*3), + retry.Settings{ + OnError: func(_ time.Duration, attempt uint64, err, lastErr error) { + if lastErr == nil || err.Error() != lastErr.Error() { + log := h.logger.Debugw + if attempt > 2 { + log = h.logger.Infow + } + + log("Can't update or insert instance. Retrying", zap.Error(err), zap.Uint64("retry count", attempt)) + } + }, + }, + ) + if err != nil { + return err + } + + if takeover { + // Insert the environment after each heartbeat takeover if it does not already exist in the database + // as the environment may have changed, although this is likely to happen very rarely. + if err := h.insertEnvironment(); err != nil { + return errors.Wrap(err, "can't insert environment") + } + + h.signalTakeover() + } else if otherResponsible { + if state, _ := h.state.Load(); !state.otherResponsible { + state.otherResponsible = true + h.state.Store(state) + } + } + + return nil +} + +func (h *HA) realizeLostHeartbeat() { + stmt := h.db.Rebind("UPDATE icingadb_instance SET responsible = ? WHERE id = ?") + if _, err := h.db.ExecContext(h.ctx, stmt, "n", h.instanceId); err != nil && !utils.IsContextCanceled(err) { + h.logger.Warnw("Can't update instance", zap.Error(internal.CantPerformQuery(err, stmt))) + } +} + +// insertEnvironment inserts the environment from the specified state into the database if it does not already exist. +func (h *HA) insertEnvironment() error { + // Instead of checking whether the environment already exists, use an INSERT statement that does nothing if it does. + stmt, _ := h.db.BuildInsertIgnoreStmt(h.environment) + + if _, err := h.db.NamedExecContext(h.ctx, stmt, h.environment); err != nil { + return internal.CantPerformQuery(err, stmt) + } + + return nil +} + +func (h *HA) removeInstance(ctx context.Context) { + h.logger.Debugw("Removing our row from icingadb_instance", zap.String("instance_id", hex.EncodeToString(h.instanceId))) + // Intentionally not using h.ctx here as it's already cancelled. + query := h.db.Rebind("DELETE FROM icingadb_instance WHERE id = ?") + _, err := h.db.ExecContext(ctx, query, h.instanceId) + if err != nil { + h.logger.Warnw("Could not remove instance from database", zap.Error(err), zap.String("query", query)) + } +} + +func (h *HA) removeOldInstances(s *icingaredisv1.IcingaStatus, envId types.Binary) { + select { + case <-h.ctx.Done(): + return + case <-time.After(timeout): + query := h.db.Rebind("DELETE FROM icingadb_instance " + + "WHERE id <> ? AND environment_id = ? AND endpoint_id = ? AND heartbeat < ?") + heartbeat := types.UnixMilli(time.Now().Add(-timeout)) + result, err := h.db.ExecContext(h.ctx, query, h.instanceId, envId, + s.EndpointId, heartbeat) + if err != nil { + h.logger.Errorw("Can't remove rows of old instances", zap.Error(err), + zap.String("query", query), + zap.String("id", h.instanceId.String()), zap.String("environment_id", envId.String()), + zap.String("endpoint_id", s.EndpointId.String()), zap.Time("heartbeat", heartbeat.Time())) + return + } + affected, err := result.RowsAffected() + if err != nil { + h.logger.Errorw("Can't get number of removed old instances", zap.Error(err)) + return + } + h.logger.Debugf("Removed %d old instances", affected) + } +} + +func (h *HA) signalHandover() { + if h.responsible { + h.state.Store(haState{ + responsibleTsMilli: time.Now().UnixMilli(), + responsible: false, + otherResponsible: false, + }) + + select { + case h.handover <- struct{}{}: + h.responsible = false + case <-h.ctx.Done(): + // Noop + } + } +} + +func (h *HA) signalTakeover() { + if !h.responsible { + h.state.Store(haState{ + responsibleTsMilli: time.Now().UnixMilli(), + responsible: true, + otherResponsible: false, + }) + + select { + case h.takeover <- struct{}{}: + h.responsible = true + case <-h.ctx.Done(): + // Noop + } + } +} |