-
Notifications
You must be signed in to change notification settings - Fork 17
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
Enable --remote_download_minimal #209
Enable --remote_download_minimal #209
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I added the equivalent mac-arm jobs since those are cached, this PR build is already finished in 3 minutes and 43 seconds. Basline master is 15 minutes and 33 seconds. I wasn't sure where you were looking to determine where seconds in bazel
and seconds in prep / cleanup
go but it seems like we should let this run and revert if we start encountering problems.
I didn't fully understand some of the issues contributors on the bazel issue were describing, but it seemed like you can only run into problems if things start getting deleted from the cache somehow. I don't think we will encounter that since nightly and continuous just overwrite everything? We just use --remote_upload_local_results=yes
, there doesn't appear to be anything like "remote upload minimal" at least.
And just tell the buildcops on slack if they start seeing errors like java.lang.IllegalStateException: Unexpected inconsistency: ActionLookupData
or java.io.IOException: Failed to fetch file with hash
then we'll probably want to revert this.
Reviewed 1 of 1 files at r1, all commit messages.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @jwnimmer-tri)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Reviewed 1 of 1 files at r1, all commit messages.
Reviewable status: complete! all files reviewed, all discussions resolved (waiting on @jwnimmer-tri)
Maybe. It's not so much "overwrite" as "repopulate" though. I think the risk is that "access time" is no longer a sound measure of what is still relevant in the cache, so our "evict-after-a-few-days-without-access" script rule might delete a build output, but not delete the ac entry that refers to that output. I also realized -- if we see trouble here, we might need to not only revert this patch but also bump the cache key version from v2 => v3 to force a re-population. drake-ci/driver/configurations/cache.cmake Line 493 in fa06c03
|
Baseline from master:
Canary using this pull request:
This risks bazelbuild/bazel#8250 if our cache TTL is too low, but I am thinking we're probably safe.
This change is