You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Recently there is a spike in ingest elasticsearch test failure due to operation timeout, e.g., https://github.com/Unstructured-IO/unstructured/actions/runs/6372547014/job/17296165138 even after we have started to allow for retries.
Exploration suggest that add more timeout allowance and more retry allowance can mitigate this problem, which suggests the root issue is CI resource being tight.
We should profile the ingest CI to track resource usage for memory, CPU time, and storage (since a lot of test also creates local files) and investigate where the hotspots are. Specifically for this ticket right before elasticsearch test starts and after es host is created. If the root issue is indeed resource being too small we should followup to investigate if the problem is tests not releasing resource properly or the demand of a single ingest test can be larger than the CI instance allocation.
The text was updated successfully, but these errors were encountered:
Recently there is a spike in ingest elasticsearch test failure due to operation timeout, e.g., https://github.com/Unstructured-IO/unstructured/actions/runs/6372547014/job/17296165138 even after we have started to allow for retries.
Exploration suggest that add more timeout allowance and more retry allowance can mitigate this problem, which suggests the root issue is CI resource being tight.
We should profile the ingest CI to track resource usage for memory, CPU time, and storage (since a lot of test also creates local files) and investigate where the hotspots are. Specifically for this ticket right before elasticsearch test starts and after es host is created. If the root issue is indeed resource being too small we should followup to investigate if the problem is tests not releasing resource properly or the demand of a single ingest test can be larger than the CI instance allocation.
The text was updated successfully, but these errors were encountered: