fix: allow artifact gc to delete directory. Fixes #12857 #13091
Conversation
Signed-off-by: Tianchu Zhao <evantczhao@gmail.com>
tczhao
changed the title
tczhao
changed the title
workflow/artifacts/s3/s3.go
Outdated
| isDir, err := s3cli.IsDirectory(artifact.S3.Bucket, artifact.S3.Key) | ||
| if err != nil { | ||
| return fmt.Errorf("failed to test if %s is a directory: %v", artifact.S3.Key, err) | ||
| } | ||
|
|
||
| if !isDir { | ||
| return s3cli.Delete(artifact.S3.Bucket, artifact.S3.Key) | ||
| } else { |
There was a problem hiding this comment.
this could potentially be optimized I'm thinking -- try deleting and if it doesn't work and gives the appropriate error, then proceed with the directory deletion
similar to what I did in #12974
There was a problem hiding this comment.
Make sense, the current change doubles the API calls
There was a problem hiding this comment.
However, looking deeper I realise
For
err = deleteFile
if isNotFound(err) {
deleteDir
}to work, deleteFile needs to return the appropriate error.
In s3, delete only adds a marker and will return 204 regardless if the key exist or not
There was a problem hiding this comment.
Updated, use simple path suffix check instead of s3cli.ListDirectory to determine if the path is a directory
There was a problem hiding this comment.
Huh does the / check actually work? Great workaround if so!
workflow/artifacts/s3/s3.go
Outdated
| for _, objKey := range keys { | ||
| err = s3cli.Delete(artifact.S3.Bucket, objKey) |
There was a problem hiding this comment.
this can probably be parallelized.
I am also a bit concerned that for large directories this could potentially hit a rate limit or cause a lot of load, although I think in this case archive: none is already an edge case and a large dir would be another edge case, so potentially not necessary to think about that. though maybe should leave a comment for future reference
There was a problem hiding this comment.
err := retry.OnError(retry.DefaultBackoff, isTransientS3Err, func() error {
currently we have retry.DefaultBackoff and isTransientS3Err to handle the rate limit situation
There was a problem hiding this comment.
Ah right above this code/surrounding it, good point.
I do think this could still be parallelized, although that can be a separate PR as it seems we lack parallelization in several (most?) places for artifacts (related: #12442)
agilgur5
changed the title
agilgur5
added
the
area/gc
Co-authored-by: Anton Gilgur <4970083+agilgur5@users.noreply.github.com> Signed-off-by: Tianchu Zhao <evantczhao@gmail.com>
Co-authored-by: Anton Gilgur <4970083+agilgur5@users.noreply.github.com> Signed-off-by: Tianchu Zhao <evantczhao@gmail.com>
Signed-off-by: Tianchu Zhao <evantczhao@gmail.com>
Signed-off-by: Tianchu Zhao <evantczhao@gmail.com>
Signed-off-by: Anton Gilgur <4970083+agilgur5@users.noreply.github.com> Signed-off-by: Anton Gilgur <4970083+agilgur5@users.noreply.github.com>
There was a problem hiding this comment.
LGTM if the suffix check works. It might be good to have a test case for archive: none as well
I'll leave it up to you if you want to do parallelization in this PR or separately
|
@tczhao do you want to re-trigger the CI, and then it sounds like this can be merged? |
|
CI doesn't need to be retrigerred as only the Windows tests are failing which are not required to pass. Can just hit merge |
ahh, thanks |
Signed-off-by: Tianchu Zhao <evantczhao@gmail.com> Signed-off-by: Anton Gilgur <4970083+agilgur5@users.noreply.github.com> Co-authored-by: Anton Gilgur <4970083+agilgur5@users.noreply.github.com> (cherry picked from commit a929c8f)
Fixes #12857
Motivation
We see error
The specified key does not existwhen gc artifact are not compressed and is directoryModifications
Determine if artifact is dir and delete accordingly (used in artifact gc)
Verification
Test log