Repository navigation
acc: retry postgres get/list reads in Lakebase resource tests - #7844
Conversation
gcp: TestAccept/bundle/resources/postgres_branches/update_protected TestAccept/bundle/resources/postgres_databases/update
|
Commit: Regression test reportSelected 20 changed tests (limit=20, 21 not selected) 36 tests: regression=0 unreleased=0 coverage=36 failing=0
|
✅ Integration test reportCommit: 42406a9
Top 3 slowest tests (at least 2 minutes):
|
There was a problem hiding this comment.
Automated review at fd56237. Each finding was verified against the code.
Not ready yet: part of the automated review did not complete.
This PR wraps every $CLI postgres get-* / list-* read in the Lakebase acceptance test scripts with the existing retry helper, leaving expected-failure reads (musterr, errcode) untouched. The only golden-output changes are the retry prefix in the traced command lines. No production code is modified. The change is mechanical and consistent across all 41 scripts; I found no defects to flag.
Sankalp-Mittal
left a comment
There was a problem hiding this comment.
[AI Generated] Review by review-factory (Claude Fable 5.2 + Claude Opus 5.5, cross-verified). Only points the reviewers agreed on are posted.
No blocking or major issues found.
…etry for standalone factcheck runs cloud: TestAccept/bundle/resources/postgres_projects/recreate
Changes
Wrap the
databricks postgres get-*/list-*reads in the Lakebase (postgres_*) acceptance scripts with the existingretryhelper (acceptance/bin/retry.py), as the dashboards tests already do forlakeview get. Reads that are expected to fail (musterr,errcode) are left alone. Goldens change only by theretryprefix in the traced command line.Why
On main, plain
postgres get-branch/get-databasecalls in these scripts failed on GCP with a transient Lakebase 500:Error: The request could not be processed. Please try again later and contact Databricks if the problem persists. [TraceId: [TRACE_ID]]The SDK retries only 429 and 503 (plus some message patterns). Bundle code retries the other 5xx itself (
libs/retry), but a plain CLI command run from the test script does not, so one 500 fails the whole test. The request recorder already drops a repeated failed attempt fromout.requests.txt, so the retried read keeps request goldens stable.Decisions:
retrykeeps its default of 5 attempts, 500ms apart.Tests
go test ./acceptance -run 'TestAccept/bundle/resources/postgres_|TestAccept/bundle/deployment/bind/postgres' -update; the only golden diffs are theretryprefixes.TestAccept/bundle/resources/postgres_databases/update/DMS=truepasses on AWS locally (testme-aws). In this PR's cloud run, all Lakebase tests pass on AWS, Azure and GCP (Linux and Windows), including the two that failed on main.Dashboard
This pull request and its description were written by Isaac.