Skip to content

crashActor now tears the actor down before it marks it CRASHED - #1953

Open
Luiz Oliveira (laoj2) wants to merge 1 commit into
agent-substrate:mainfrom
laoj2:call-terminate
Open

Luiz Oliveira (laoj2) wants to merge 1 commit into
agent-substrate:mainfrom
laoj2:call-terminate

Conversation

@laoj2

@laoj2 Luiz Oliveira (laoj2) commented Sep 28, 2026 •

Copy link
Copy Markdown
Collaborator

crashActor now tears the actor down before it marks it CRASHED, in this order:

  1. Call atelet.Terminate on the sandbox. This is skipped when the actor has no node, and a workload that is already gone counts as terminated.
  2. Unmount and detach the actor's external volumes.
  3. Release the worker.

If any step fails, the actor is still marked CRASHED but keeps its WorkerAssignment, so it isn't handed to another actor while a sandbox or mount may still be live on it.

  • So a retry (or RevertActor / DeleteActor) finishes the teardown and frees the worker. This also changes what a failed release does. It used to return an error and leave the actor in its previous state. Now it crashes the actor and keeps the assignment.

@cloudonly cloudonly mentioned this pull request Sep 29, 2026
2 tasks done
@laoj2
Luiz Oliveira (laoj2) force-pushed the call-terminate branch 2 times, most recently from a07ed8d to 9c27468 Compare September 29, 2026 18:02
@laoj2
Luiz Oliveira (laoj2) marked this pull request as draft October 2, 2026 17:37
@laoj2
Luiz Oliveira (laoj2) marked this pull request as ready for review October 8, 2026 21:19
@laoj2 Luiz Oliveira (laoj2) changed the title Call atelet.Terminate and detach volumes before crashing an actor crashActor now tears the actor down before it marks it CRASHED Oct 8, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

area/api-machinery area/node kind/bug Something isn't working / bugfixes

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants