Problem with Restyler Job #2742861 #194
Comments
|
Hi there, thanks for the report! The 137 exit code indicates out of memory. In other words, There are 3 possible causes for this:
With all that in mind: are you sure the |
|
We're investigating not restyling the zzz_generated files; there are some issues around that which need to be sorted out on our end before we can do that. In the meantime, I'm happy to do the measurement for item 1 if you tell me the exact thing you want measured. For item 2, if you tell me what option in For item 3, the PRs involved are changing the minimal possible number of files already: they are renaming a type and doing a search/replace of the type name in the places where it's used. |
In an ideal would, I'd like to see something like the following filled out,
I'm not exactly sure what "something small/large" is, or how to measure the maximum memory used by a process. This is kind of an open-ended debugging we're going into together here.
Unfortunately, on second look it's not as easy as I thought to do that. Let me think about this one a bit and see if I can do some work to enable that. I'm curious to learn more about how this is blocking. If you've run the restyle locally and are sure Restyled wouldn't change anything it were to succeed, can you not just ignore the error for this particular PR? |
|
For what it's worth, I am now seeing the error on project-chip/connectedhomeip#24398 where last week that same PR (with very minor changes) was passing.
OK, so you just want the memory used by clang-format itself, not the entire docker image? And I assume you want this for clang-format 9, which is what the restyler is running? That might be a little complicated; clang-format 9 is a bit of a pain to get your hands on (at least on Mac, which is what I have to work with).
No, because the restyle job is marked as a required CI job for the relevant repository. I'd have to find one of the admins to admin-merge it, which is ... a lot of overhead. And I am getting this for 5 different PRs at this point, and likely more to come. |
Either is fine; though if you found a marked difference between the two that would be useful information! As I mentioned, the docker image does absolutely nothing besides run
I see. Well I'm honored to be a required status! As a workaround, you could disable
Well this is definitely interesting and points to something systemic that is wrong with the underlying instances where this stuff runs. But that's unlikely since natural scaling activities have replaced the instances multiple times, as recently as 5PM ET. I'll have to continue digging. |
Wait, I misread a graph! The current instance has been around a while (though it certainly seems healthy). I'll rotate it out and see if that helps. |
|
For what it's worth, at this point we've stopped restyling the zzz_generated files, so that issue is addressed. That said, when I run restyled locally on an Intel Mac on the set of files changed in project-chip/connectedhomeip@52b1eb0 and project-chip/connectedhomeip@4abb1f0 and project-chip/connectedhomeip@47e37c9 (which is the relevant set for the failure above) I see the docker claiming that the memory for the image goes up to 490MB (it's sampling, I guess, so not sure whether it goes higher). So it's pretty borderline if 512MB is the cap.... |
|
What command do you use to run restyled locally? |
|
Based on my metrics, things seem better since I rotated the instance. It's also possible you just aren't running many jobs yet (it's early in ET at least). |
We also stopped restyling the generated files... |
|
I you remove Are things looking better on your end? I'm not showing any failures in your project since 7:40PM ET last night, which is basically when I rotated the instance. |
That doesn't seem to fail, on a set of files that CI failed on.... On our end, we are no longer restyling the generated files, so I would not expect this to come up as an issue much anymore.... We can probably just close this. |
Sounds good. Based on the evidence so far, it seems like something was going on with the underlying EC2 instance and rotating it out was the solution. Skipping generated files is still a Good Thing, generally speaking though, to avoid these sorts of issues when they're not caused by something systemic. |
Please use the below template to report an issue with this Job. Don't submit
sensitive information, this is a public project. If you'd rather communicate
privately, email [email protected].
Hi there-
I'm having a problem with a Restyled Job
What I expected to happen: The job should pass, because I already locally ran the restyled.io docker bits on my code.
What happened instead: The job failed with:
Configuration (if applicable):
The text was updated successfully, but these errors were encountered: