Skip to content

Conversation

@vrothberg
Copy link
Member

Avoid redundantly pulling the same image (or layer).

Signed-off-by: Valentin Rothberg rothberg@redhat.com

@vrothberg
Copy link
Member Author

Do not merge this PR, it's used to increase confidence for the changes to storage and image by letting it pass Buildah's CI.

@vrothberg
Copy link
Member Author

This PR needs to be updated. We also need to merge #1400 first to make it pass.

@rh-atomic-bot
Copy link
Collaborator

☔ The latest upstream changes (presumably 3d74031) made this pull request unmergeable. Please resolve the merge conflicts.

@rhatdan
Copy link
Member

rhatdan commented Mar 27, 2019

@vrothberg #1400 Is merged can you rework this one?

@vrothberg
Copy link
Member Author

@vrothberg #1400 Is merged can you rework this one?

Sure. Already rebased and pushed.

@rhatdan
Copy link
Member

rhatdan commented Mar 27, 2019

Can you remove [WIP]?
LGTM

@vrothberg
Copy link
Member Author

Can you remove [WIP]?

Not until the PRs over image, storage and skopeo are merged. This PR is testing the new copy-detection mechanism.

I can open a dedicated PR to vendor in the latest image/storage.

@vrothberg
Copy link
Member Author

All tests are hitting a timeout. I'll vendor the storage changes first to exclude some potential error sources from image.

@vrothberg vrothberg force-pushed the copy-detection branch 7 times, most recently from e306618 to 6ee4f15 Compare March 28, 2019 12:44
@vrothberg
Copy link
Member Author

@mtrmac @nalind @TomSweeneyRedHat the integration tests don't seem to fire off as the last logs are calling tests/test_runner.sh and do not show any log afterward. containers/podman#2763 over at libpod is passing now. I try to wrap my head around it I fail to reproduce.

@vrothberg
Copy link
Member Author

It ultimately suggests a deadlock but I don't see how.

@vrothberg
Copy link
Member Author

Good news before the weekend. No deadlock and tests are passing now. I took a shot a travis_wait which is famous for suppressing output when a timeout is being hit.

I don't yet fully understand WHY we hit the timeout but removing travis_wait let's us pass again. One really annoying thing about travis_wait is that it does not print any log at the time out which complicates debugging.

I've already checked with @TomSweeneyRedHat who wants a head nod from @nalind to make sure that removing travis_wait does not introduce any undesired side-effect.

}

func (s *blobCacheDestination) SupportsBlobLocks() bool {
return false
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

(Note that the cache object can wrap a storageImageDestination in Commit. I don’t quite think that it matters — we are only actually creating some new layers, so there should not be any concurrent downloads of that.)

Avoid redundantly pulling the same image (or layer).

Signed-off-by: Valentin Rothberg <rothberg@redhat.com>
@rh-atomic-bot
Copy link
Collaborator

☔ The latest upstream changes (presumably fcc12bd) made this pull request unmergeable. Please resolve the merge conflicts.

@vrothberg
Copy link
Member Author

vrothberg commented Apr 18, 2019

Closing in favour of containers/image#611.

@vrothberg vrothberg closed this Apr 18, 2019
@github-actions github-actions bot locked as resolved and limited conversation to collaborators Oct 13, 2023
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants