THREESCALE-11940: Remove transient flag from the fetcher task #441
Conversation
|
|
||
| def start_to_fetch_jobs | ||
| Async(transient: true) { @job_fetcher.start(@jobs) } | ||
| Async { @job_fetcher.start(@jobs) } |
There was a problem hiding this comment.
Transient means what? That this block is detached from it's immediate parent? Or AI was lying to me?
There was a problem hiding this comment.
https://socketry.github.io/async/guides/tasks/index.html#reactor-lifecycle
This was an idea to investigate the CrashLoopBack the client is having. This is the hypotetical scenario to test:
- Some error happens somewhere, that leads the worker to shutdown.
- The Job fetcher task is still working on the background, but the main task won't wait for it because is transient.
- As part of the shutdown, at some point the logger is closed.
- Then the job fetcher task fails. That makes it try to log, and we see the errors about the log being closed
So if we don't mark the task as transient, the main task will wait for it and we hopefully will see any error logged.
This is just an hypothesis and proving it would require creating an image and instruct the client to run it. So I'm trying other approaches for now. See my comments in the issue.
There was a problem hiding this comment.
You're saying that a transient task will not block process exit?
There was a problem hiding this comment.
I don't think it would block the process to exit, the docs say the transient task is forcefully stopped then the reactor it belongs to finishes. But I don't know, maybe at some point the transient task is still running when it shouldn't. Just speculating, as the Jira issue is damn complicated.
There was a problem hiding this comment.
Another scenario: The process_all method could fail or exit so fast that the fetch still haven't started to fetch, if that happens, the worker would start a shutdown and the fetcher wouldn't prevent it because it's transient.
2 participants
Just to make some tests. Ignore for now
Issue
THREESCALE-11940