fix: ignore errors to stop pods

If Talos fails to stop the pods, it doesn't affect any future operations. Prevously it was critical for the install/upgrade to succeed, as it tried to remove whole partition table, but now even dangling mounts won't affect upgrades. Sometimes pods can't be stopped because of network file systems. Fixes #10174 Signed-off-by: Andrey Smirnov <[email protected]>
siderolabs · Feb 11, 2025 · 711cf2d · 711cf2d
1 parent 142d754
commit 711cf2d
Showing 1 changed file with 6 additions and 2 deletions.
diff --git a/internal/app/machined/pkg/runtime/v1alpha1/v1alpha1_sequencer_tasks.go b/internal/app/machined/pkg/runtime/v1alpha1/v1alpha1_sequencer_tasks.go
@@ -1230,13 +1230,17 @@ func stopAndRemoveAllPods(stopAction cri.StopAction) runtime.TaskExecutionFunc {
 		// CNI, preventing the CRI from cleaning up the pod's networking.
 
 		if err = client.StopAndRemovePodSandboxes(ctx, stopAction, runtimeapi.NamespaceMode_POD, runtimeapi.NamespaceMode_CONTAINER); err != nil {
-			return err
+			logger.Printf("failed to stop and remove pods with POD network mode: %s", err)
 		}
 
 		// With the POD network mode pods out of the way, we kill the remaining
 		// pods.
 
-		return client.StopAndRemovePodSandboxes(ctx, stopAction)
+		if err = client.StopAndRemovePodSandboxes(ctx, stopAction); err != nil {
+			logger.Printf("failed to stop and remove pods: %s", err)
+		}
+
+		return nil
 	}
 }