Method failoverLocked

pkg/api/cluster_client.go:159–202

failoverLocked performs the failover while holding the write lock.

(ctx context.Context)

// failoverLocked performs the failover while holding the write lock.
func (cc *ClusterClient) failoverLocked(ctx context.Context) error {
	if len(cc.candidates) == 0 {
		return fmt.Errorf("no candidates available for failover in cluster group '%s'", cc.groupName)
	}

	oldProfile := cc.activeProfile
	numCandidates := len(cc.candidates)

	cc.logger.Info("[CLUSTER] Initiating failover from %s", oldProfile)

	// Try each candidate starting from the next one in order
	for attempt := 1; attempt <= numCandidates; attempt++ {
		idx := (cc.activeIndex + attempt) % numCandidates
		entry := cc.candidates[idx]

		cc.logger.Info("[CLUSTER] Failover attempt %d/%d: trying %s",
			attempt, numCandidates, entry.Name)

		client, err := NewClient(entry.Config,
			WithLogger(cc.logger),
			WithCache(cc.cache),
		)
		if err != nil {
			cc.logger.Error("[CLUSTER] Failover to %s failed: %v", entry.Name, err)
			continue
		}

		cc.activeClient = client
		cc.activeProfile = entry.Name
		cc.activeIndex = idx
		cc.logger.Info("[CLUSTER] Failover successful: %s -> %s", oldProfile, entry.Name)

		// Notify callback outside the lock to avoid deadlocks
		if cc.onFailover != nil {
			callback := cc.onFailover
			newProfile := entry.Name
			go callback(oldProfile, newProfile)
		}

		return nil
	}

	return fmt.Errorf("failover exhausted: no reachable candidate in cluster group '%s'", cc.groupName)
}

// StartHealthCheck begins periodic health checking of the active node.
// If the active node becomes unreachable, it automatically triggers failover.