vitessio/vitess · error

FindReplicas: LookupHost failed

Error message

FindReplicas: LookupHost failed %v

What it means

After splitting a replica address into host and port, FindReplicas performs net.LookupHost(host) to resolve the hostname to IPs. DNS resolution failure produces this error and aborts the lookup of all replica IPs.

Solutions

1. Verify `getent hosts <replica-host>` (or nslookup) resolves on the machine running FindReplicas
2. Fix DNS configuration or add the replica hostnames to /etc/hosts
3. Update replication report_host settings to use resolvable names or IPs
4. Retry after DNS recovers; if transient, consider caching/resilient resolution upstream

Example fix

// before: report_host = replica-1.internal (unresolvable)
// after: in replica my.cnf
// report_host = 10.0.0.5  // or a DNS-resolvable name

// pre-check resolvability of replica hosts
host := "replica-1.internal"
if ips, err := net.LookupHost(host); err != nil {
    log.Warnf("host %s not resolvable: %v", host, err)
} else {
    _ = ips
}

addrs, err := mysqld.FindReplicas()
if err != nil && strings.Contains(err.Error(), "LookupHost failed") {
    time.Sleep(2 * time.Second) // transient DNS
    addrs, err = mysqld.FindReplicas()
}

When it happens

Trigger: Calling FindReplicas/GetReplicas when a replica hostname from replication status cannot be resolved by DNS — e.g. stale DNS entries, replicas behind NAT, or hostnames absent from /etc/hosts.

Common situations: Kubernetes/container environments where pod hostnames change; DNS outages or misconfigured resolv.conf; replication configured with hostnames that no longer exist.

Related errors

• FindReplicas: malformed addr
• can't dial mysqlctld
• Connection DNS for has changed; old: new
• Connection DNS for reporting as empty, ignoring
• err

AI-assisted analysis of vitessio/vitess@01a25a7d17 (2026-09-01). Data as JSON: /api/errors/3458b85c482daf66. Report an issue: GitHub.

Appendix: source

Thrown at go/vt/mysqlctl/replication.go:1164

	for _, row := range qr.Rows {
		// Check for prefix, since it could be "Binlog Dump GTID".
		if strings.HasPrefix(row[colCommand].ToString(), binlogDumpCommand) {
			host := row[colClientAddr].ToString()
			if host == "localhost" {
				// If we have a local binlog streamer, it will
				// show up as being connected
				// from 'localhost' through the local
				// socket. Ignore it.
				continue
			}
			host, _, err = netutil.SplitHostPort(host)
			if err != nil {
				return nil, fmt.Errorf("FindReplicas: malformed addr %v", err)
			}
			var ips []string
			ips, err = net.LookupHost(host)
			if err != nil {
				return nil, fmt.Errorf("FindReplicas: LookupHost failed %v", err)
			}
			addrs = append(addrs, ips...)
		}
	}

	return addrs, nil
}

// GetGTIDMode gets the GTID mode for the server
func (mysqld *Mysqld) GetGTIDMode(ctx context.Context) (string, error) {
	conn, err := getPoolReconnect(ctx, mysqld.dbaPool)
	if err != nil {
		return "", err
	}
	defer conn.Recycle()

	return conn.Conn.GetGTIDMode()
}

View on GitHub (pinned to 01a25a7d17)