-
Notifications
You must be signed in to change notification settings - Fork 51
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Prints "Could not find any config for exited child..." upon upgrade #46
Comments
(As a note, looks like this can also be triggered when |
Hmm. So my recollection is that I wrote the code in #39 because our einhorn worker monitoring would fire because einhorn would keep the state passer in its state variables indefinitely, thus looking like a very old worker process. But in any case, it feels like the code in #40 should safely handle all possible cases of this race - it's basically a more generalized and better version of the same issues as #39. Maybe we can revert 673f5a8 now that #40 is in place? We should make sure we think through the state upgrade path before we do, though. I'm not totally confident all of that is right, though. But intuitively, it seems like a better fix than making the communication with the state passer more complicated. |
Ah yeah, I like that much better. I'll poke at it.
On Sun, Apr 12, 2015 at 11:41 PM, Evan Broder notifications@github.com
|
Did you guys ever figure out how to resolve this? |
I now see this pretty consistently:
Looks like it was introduced in: 673f5a8. Sounds like we had a race before, but I think this means we have a race still.
@ebroder, what would you think about de-racing by making the state_passer live past the SIGCHLD handler install? I'd guess that would mean either (a) installing the SIGCHLD handle before loading state — though I haven't checked if there's a chicken and an egg there — or (b) having some coordination mechanism between the state_passer and the master. Thoughts?
The text was updated successfully, but these errors were encountered: