Merge pull request #47044 from corhere/lazy-health-status

Lazily checkpoint container health status to disk

container/container.go

@@ -267,6 +267,32 @@ func (container *Container) WriteHostConfig() (*containertypes.HostConfig, error
 	return &deepCopy, nil
 }
 
+// CommitInMemory makes the Container's current state visible to queries,
+// but does not persist state.
+//
+// Callers must hold a Container lock.
+func (container *Container) CommitInMemory(store *ViewDB) error {
+	var buf bytes.Buffer
+	if err := json.NewEncoder(&buf).Encode(container); err != nil {
+		return err
+	}
+
+	var deepCopy Container
+	if err := json.NewDecoder(&buf).Decode(&deepCopy); err != nil {
+		return err
+	}
+
+	buf.Reset()
+	if err := json.NewEncoder(&buf).Encode(container.HostConfig); err != nil {
+		return err
+	}
+	if err := json.NewDecoder(&buf).Decode(&deepCopy.HostConfig); err != nil {
+		return err
+	}
+
+	return store.Save(&deepCopy)
+}
+
 // SetupWorkingDirectory sets up the container's working directory as set in container.Config.WorkingDir
 func (container *Container) SetupWorkingDirectory(rootIdentity idtools.Identity) error {
 	if container.Config.WorkingDir == "" {

daemon/health.go

@@ -231,8 +231,11 @@ func handleProbeResult(d *Daemon, c *container.Container, result *types.Healthch
 		// Else we're starting or healthy. Stay in that state.
 	}
 
-	// replicate Health status changes
-	if err := c.CheckpointTo(d.containersReplica); err != nil {
+	// Replicate Health status changes to the API, skipping persistent storage
+	// to avoid unnecessary disk writes. The health state is only best-effort
+	// persisted across of the daemon. It will get written to disk on the next
+	// checkpoint, such as when the container state changes.
+	if err := c.CommitInMemory(d.containersReplica); err != nil {
 		// queries will be inconsistent until the next probe runs or other state mutations
 		// checkpoint the container
 		log.G(context.TODO()).Errorf("Error replicating health state for container %s: %v", c.ID, err)