Failing on Creating MatchingIndexEngine - Can't find logs to debug.

[MegilFGallant]

I'm working on creating an index via the vertex ai matching engine but the creation job is failing and I'm not able to find the logs it's pointing me to in the log explorer. I'd like to fix the issue with the records before rerunning since it takes 20 minutes to run before failure.

here is the error:

FailedPrecondition: 400 There are invalid records under files [gs://vertexai_embedding_artifacts/matching_engine/initial/embedding_data.json]. Please check the details in the metadata of operation: projects//locations/us-central1/indexes/1256425131199168512/operations/5606239022130135040. 9: There are invalid records under files [gs://vertexai_embedding_artifacts/matching_engine/initial/embedding_data.json]. Please check the details in the metadata of operation: projects//locations/us-central1/indexes/1256425131199168512/operations/5606239022130135040.

[sararob]

In the Vertex AI Matching Engine section of your Cloud console, you can click on the index you're creating and there is a "View logs" link in the top right.

[rjimr]

In the Vertex AI Matching Engine section of your Cloud console, you can click on the index you're creating and there is a "View logs" link in the top right.

I don't see such an option available on the failed index. Perhaps, this option will be only against the successfully created indexes.

[sararob]

Can you try going to the Logs Explorer in your console and searching for the Operation ID in the query box?

[echa-cedar]

Hi, I'm facing the same problem - I'm unable to "click" on the failed index. And when searching operations ID in the Logs Explorer, there is nothing that is showing..

[chris-ng-scmp]

Got the same issue No way to know why indexes are Creation failed

And no log logged in Logs Explorer at all for the index creation

[Ludovic-Emo-Pyl-Tech]

Hi,

We have got the same issue here, could a solution be implemented for the UI ?

In the meantime, as a workaround, we can use the following gcloud command however it would be more efficient to read the reason in the UI.

gcloud ai operations describe xxx

where xxx is the id found within the error message:
[...] Please check the details in the metadata of operation: projects/zzz/locations/yyy/indexes/iii/operations/xxx [...]

in our case it was: errorMessage: 'Embedding size mismatch: expect 768 but get 767'

beginner error, easy to fix, but easier to fix with the error message ;)

Thanks in advance for your help.