I'm out today and tomorrow so won't be able to help look at this, but I will note at least that we didn't land any of the changes to limit the number of files in the archives, so that might be a part of the problem if archive got a lot slower.

It's the UKM stderr thing again.

Suspected CL is https://chromium-review.googlesource.com/c/chromium/src/+/757885. Revert is in the CQ.

stderr contains the following:

[5416:2284:1109/072808.502:ERROR:service_manager.cc(158)] Connection InterfaceProviderSpec prevented service: content_renderer from binding interface: ukm::mojom::UkmRecorderInterface exposed by: content_browser
[5416:2284:1109/072808.714:ERROR:service_manager.cc(158)] Connection InterfaceProviderSpec prevented service: content_renderer from binding interface: ukm::mojom::UkmRecorderInterface exposed by: content_browser

Hi, John.

Do you think this is likely to be the same issue as  crbug.com/783247  "archive_webkit_tests_results failing on multiple builders"?

Thanks!

I am curious: could a change like 757885 affect "archive_webkit_tests_results"? I naively think that archiving step doesn't run any layout tests, instead it just zips things and stores the result somewhere.

Yeah, it results in a massively increased number of files that we have to archive. The same thing happened in https://chromium-review.googlesource.com/c/chromium/src/+/722301, the previous attempt to land the same CL.

re "massively increased": I downloaded one of the archives & see over 61000 stderr files matching #4.

Ah, that makes good sense! Thanks for explanation!

I'm really sorry and thank you for treating this. Now my culprit CL was reverted, the problem was fixed?

#11: no worries. It should be, but I'm still waiting on post-revert data to make it into our monitoring.

2hr metrics are returning to their previous levels: http://shortn/_2eOTKGYsdp