kubernetes/kops · error

already started

Error message

already started

What it means

The otlptracefile client implements the otlptrace.Client interface; Start initializes the file writer and may only be called once. If Start is invoked while c.writer is already non-nil (a previous Start succeeded), it returns "already started" to prevent re-initializing and clobbering the open trace file.

Source

Thrown at pkg/otel/otlptracefile/client.go:59

	var cfg Config
	for _, option := range opts {
		option(&cfg)
	}

	c := &client{
		cfg: cfg,
	}

	return c
}

// Start implements otlptrace.Client.
func (c *client) Start(ctx context.Context) error {
	c.writerMutex.Lock()
	defer c.writerMutex.Unlock()

	if c.writer != nil {
		return fmt.Errorf("already started")
	}

	w, err := newWriter(c.cfg)
	if err != nil {
		return err
	}
	c.writer = w

	return nil
}

// Stop implements otlptrace.Client.
func (c *client) Stop(ctx context.Context) error {
	c.writerMutex.Lock()
	defer c.writerMutex.Unlock()

	if c.writer != nil {
		err := c.writer.Close()

View on GitHub (pinned to 4c8573c808)

Solutions

1. Call Start only once per client instance; guard with a sync.Once or a started flag.
2. Call Shutdown/Stop before invoking Start again, or create a fresh client instance.
3. Share one tracer provider instead of creating multiple ones around the same client.

Example fix

// before
client.Start(ctx)
client.Start(ctx) // panics path: already started
// after
var once sync.Once
once.Do(func() { client.Start(ctx) })

var started bool
func startOnce(c otlptrace.Client, ctx context.Context) error {
    if started { return nil }
    started = true
    return c.Start(ctx)
}

if err := client.Start(ctx); err != nil {
    if strings.Contains(err.Error(), "already started") {
        return nil // idempotent start
    }
    return err
}

When it happens

Trigger: Calling client.Start(ctx) twice without an intervening Shutdown; registering the same otlptracefile client with two tracers that both call Start.

Common situations: Application bootstrap code that starts the exporter in two places (e.g. both main and an init path); hot-reload logic re-calling Start on config change without shutting down first.

Related errors

• already closed
• incorrect syntax for lifecyle-overrides, correct syntax is T
• unknown lifecycle %q, available lifecycle: %s
• duplicate scope: %q
• shutdown already in progress

AI-assisted analysis of kubernetes/kops@4c8573c808 (2026-09-05). Data as JSON: /api/errors/12e16475b1c773f6. Report an issue: GitHub.