feat: add resilient container restart controls

This commit is contained in:
admin committed 2026-09-07 13:27:27 +02:00
1 parent 1bbb14470c
commit 7bb33e528c
13 files changed
+747 -28

No files matched your search

+79 -2
View File
@@ -64,7 +64,7 @@ func (r *DockerRunner) Update(ctx context.Context, request updateprotocol.Reques
if !updateprotocol.ValidVersion(request.Version) {
return errors.New("invalid target version")
}
if err := r.docker.negotiate(ctx); err != nil {
if err := waitForDockerAPI(ctx, 30*time.Second, time.Second, r.docker.negotiate); err != nil {
return &updateError{message: "Containerverwaltung ist nicht erreichbar; der Dienst läuft unverändert weiter", err: err}
}
target, err := r.docker.findTarget(ctx, r.targetLabel)
@@ -75,7 +75,18 @@ func (r *DockerRunner) Update(ctx context.Context, request updateprotocol.Reques
if err != nil {
return err
}
if inspected.State.Health == nil || inspected.State.Health.Status != "healthy" {
if inspected.State.Health == nil {
return &updateError{message: "Die laufende Instanz ist nicht healthy; das Update wurde sicherheitshalber nicht gestartet", err: errors.New("current container is not healthy")}
}
switch inspected.State.Health.Status {
case "healthy":
case "starting":
progress(updateprotocol.PhaseQueued, "Laufender Container beendet noch seinen Start-Healthcheck")
inspected, err = r.docker.waitCurrentHealthy(ctx, target.ID, inspected.Image, 100*time.Second, time.Second)
if err != nil {
return &updateError{message: "Die laufende Instanz wurde nach dem Host-Start nicht rechtzeitig healthy; das Update wurde nicht gestartet", err: err}
}
default:
return &updateError{message: "Die laufende Instanz ist nicht healthy; das Update wurde sicherheitshalber nicht gestartet", err: errors.New("current container is not healthy")}
}
currentVersion, err := resolveCurrentVersion(request.CurrentVersion, inspected.Config, r.imageRepository)
@@ -847,6 +858,72 @@ func (c *dockerClient) waitHealthy(ctx context.Context, id, expectedImageID, exp
}
}
func (c *dockerClient) waitCurrentHealthy(ctx context.Context, id, expectedImageID string, timeout, pollInterval time.Duration) (containerInspect, error) {
deadline := time.NewTimer(timeout)
ticker := time.NewTicker(pollInterval)
defer deadline.Stop()
defer ticker.Stop()
for {
select {
case <-ctx.Done():
return containerInspect{}, ctx.Err()
case <-deadline.C:
return containerInspect{}, errors.New("current container did not become healthy in time")
case <-ticker.C:
inspected, err := c.inspectContainer(ctx, id)
if err != nil {
return containerInspect{}, err
}
if !inspected.State.Running {
return containerInspect{}, fmt.Errorf("current container stopped with exit code %d", inspected.State.ExitCode)
}
if expectedImageID != "" && inspected.Image != expectedImageID {
return containerInspect{}, fmt.Errorf("current container image changed from %s to %s", expectedImageID, inspected.Image)
}
if inspected.State.Health == nil {
return containerInspect{}, errors.New("current container has no healthcheck")
}
switch inspected.State.Health.Status {
case "healthy":
return inspected, nil
case "starting":
continue
default:
return containerInspect{}, fmt.Errorf("current container health is %q", inspected.State.Health.Status)
}
}
}
}
func waitForDockerAPI(ctx context.Context, timeout, interval time.Duration, negotiate func(context.Context) error) error {
retryCtx, cancel := context.WithTimeout(ctx, timeout)
defer cancel()
var lastErr error
for {
attemptCtx, attemptCancel := context.WithTimeout(retryCtx, 3*time.Second)
lastErr = negotiate(attemptCtx)
attemptCancel()
if lastErr == nil {
return nil
}
timer := time.NewTimer(interval)
select {
case <-retryCtx.Done():
if !timer.Stop() {
select {
case <-timer.C:
default:
}
}
if ctx.Err() != nil {
return ctx.Err()
}
return lastErr
case <-timer.C:
}
}
}
func (c *dockerClient) request(ctx context.Context, method, path string, body io.Reader, headers http.Header, versioned bool) (*http.Response, error) {
if versioned {
path = c.apiPrefix + path