[test-dataproc] make test-dataproc-37 and -38 not race #13573
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Consider, for example, this deploy: https://ci.hail.is/batches/7956812. `test-dataproc-37` succeeded but `test-dataproc-38` failed (it timed out b/c the master failed to come online). You can see the error logs for the cluster here: https://cloudlogging.app.goo.gl/t1ux8oqy11Ba2dih7 It states a certain file either did not exist or we did not have permission to access it. [`test_dataproc-37`](https://batch.hail.is/batches/7956812/jobs/193) and [`test_dataproc-38`](https://batch.hail.is/batches/7956812/jobs/194) started around the same time and both uploaded four files into: gs://hail-30-day/hailctl/dataproc/ci_test_dataproc/0.2.121-7343e9c368dc/ And then set it to public read/write. The public read/write means that permissions are not the issue. Instead, the issue is that there must be some sort of race condition in GCS which means that if you "patch" (aka overwrite) an existing file, it is possible that a concurrent reader will see the file as not existing.
daniel-goldstein
approved these changes
Sep 6, 2023
added 2 commits
September 6, 2023 17:10
CHANGELOG: Mitigate hail-is#12936 in which VEP Dataproc clusters fail to start. The root cause is complex. Docker has a bug which prevents it from cleanly starting if it is *re* installed. Whatever Google is doing in Dataproc to configure their Docker "component" appears to trigger this bug. See for details: hail-is#12936 (comment) The basic fix is to sleep to allow the system to coalesce a bit and then to restart Docker.
|
I merged the approved PR #13580 into this. |
|
For a bit of extra confidence, I ran test-dataproc on this branch: https://ci.hail.is/batches/7960963 The clusters created successfully and the tests are running now. |
|
test-dataproc passed. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Consider, for example, this deploy: https://ci.hail.is/batches/7956812.
test-dataproc-37succeeded buttest-dataproc-38failed (it timed out b/c the master failed to come online).You can see the error logs for the cluster here: https://cloudlogging.app.goo.gl/t1ux8oqy11Ba2dih7
It states a certain file either did not exist or we did not have permission to access it.
test_dataproc-37andtest_dataproc-38started around the same time and both uploaded four files into:And then set it to public read/write. The public read/write means that permissions are not the issue.
Instead, the issue is that there must be some sort of race condition in GCS which means that if you "patch" (aka overwrite) an existing file, it is possible that a concurrent reader will see the file as not existing.
Unfortunately, I cannot confirm this with audit logs of the writes and read because public objects do not generate audit logs.