DataflowTemplates icon indicating copy to clipboard operation
DataflowTemplates copied to clipboard

[Bug]: Failing to write to temporary file storage

Open mehmetkoc182 opened this issue 2 years ago • 7 comments

Related Template(s)

BigQuery to Bigtable & Parquet File Storage to Bigtable

What happened?

There seems to be some sort of mismatch around the dataflow worker permissions and cloud storage temp directory.

I've tried 2 different templates and the issue seems to persist on the "reading" step. However, the error log seems to be related to "write" permissions I believe happens to the temporary directory for the process.

I'm using uniform bucket level access and the worker account has full access to cloud storage and cloud storage objects.

Do I need different/extra permissions for these templates?

Beam Version

Newer than 2.43.0

Relevant log output

Error initializing dynamic plugin prober" err="error (re-)creating driver directory: mkdir /usr/libexec/kubernetes: read-only file system"




ima: Can not allocate sha384 (reason: -2)

mehmetkoc182 avatar Mar 06 '23 20:03 mehmetkoc182

Is there any other error? I think that the log mentioned is "normal" to happen on GCE instances.

bvolpato avatar Mar 09 '23 02:03 bvolpato

@mehmetkoc182 Can you please elaborate / post any errors?

bvolpato avatar Mar 24 '23 14:03 bvolpato

@mehmetkoc182 Let us know if you still want to troubleshoot this. We'll close this issue if there are no further interactions.

bvolpato avatar Apr 03 '23 14:04 bvolpato

I think i have a similar error with the GCS to PubSub template, until the release 2023-03-21-00_RC00 it was working fine, but now at the 2023-03-28-00_RC00 it's failing on rename a temp file

Warning Log:

"Unable to delete temp files: "gs://my_bucket/tmp_dir/dax-tmp-2023-04-10_09_14_22-7957512643797177487-S01-0-adbd0cb3c13f0b01/[email protected]." Causes: Unable to delete file: gs://my_bucket/tmp_dir/dax-tmp-2023-04-10_09_14_22-7957512643797177487-S01-0-adbd0cb3c13f0b01/tmp-adbd0cb3c13f0932-shard--try-070f492ac9c1925b-endshard.sdfmeta."

Error Log:

Workflow failed. Causes: S01:Read Text Data/Read/Impulse+Read Text Data/Read/ParDo(OutputSingleSource)/ParMultiDo(OutputSingleSource)+Read-Text-Data-Read-ParDo-BoundedSourceAsSDFWrapper--ParMultiDo-BoundedSourceAsSDFWrapper-/PairWithRestriction+Read-Text-Data-Read-ParDo-BoundedSourceAsSDFWrapper--ParMultiDo-BoundedSourceAsSDFWrapper-/SplitWithSizing failed., Unable to rename output files from gs://my_bucket/tmp_dir/dax-tmp-2023-04-10_09_14_22-7957512643797177487-S01-0-adbd0cb3c13f0b01/[email protected] to gs://my_bucket/tmp_dir/tmp-adbd0cb3c13f0932@*.sdfmeta., Unable to rename "gs://my_bucket/tmp_dir/dax-tmp-2023-04-10_09_14_22-7957512643797177487-S01-0-adbd0cb3c13f0b01/tmp-adbd0cb3c13f0932-shard--try-070f492ac9c1925b-endshard.sdfmeta" to "gs://my_bucket/tmp_dir/tmp-adbd0cb3c13f0932-00000-of-00001.sdfmeta".

Wuerike avatar Apr 10 '23 23:04 Wuerike

@Wuerike thanks! Does the job fail? Or you see the operations being retried? Would you mind filing a support issue with the job id so the team can look at this internally?

bvolpato avatar Apr 12 '23 04:04 bvolpato

@Wuerike seen a similar case, and it was related to missing permissions on the bucket. Unsure how the update might have affected it. Can you double check on that to be sure?

bvolpato avatar Apr 18 '23 15:04 bvolpato

@bvolpato just to update, it was indeed a missing permission, I gave owner access to the service account and it worked. The problem is it was working before this is why I thought the reason could be any update.

Wuerike avatar Apr 18 '23 17:04 Wuerike

This issue has been marked as stale due to 180 days of inactivity. It will be closed in 1 week if no further activity occurs. If you think that’s incorrect or this pull request requires a review, please simply write any comment. If closed, you can revive the issue at any time. Thank you for your contributions.

github-actions[bot] avatar May 20 '24 14:05 github-actions[bot]

This issue has been closed due to lack of activity. If you think that is incorrect, or the pull request requires review, you can revive the PR at any time.

github-actions[bot] avatar May 29 '24 02:05 github-actions[bot]