test(search): de-flake nightly search-it & ui-it reindex tests

[mohityadav766]

Describe your changes:

Fixes #29187

The JavaUIIT Integration Tests (Nightly) workflow (search-it-nightly + ui-it-nightly) was failing on nearly every PR independent of content (it even failed on a no-op version-bump PR). These are timing/scope races in the #28637 reindex test suite — test-side only: no product behavior changed and no assertions weakened (two were made stricter).

Root causes & fixes

Test	Root cause	Fix
DbToEsCountReconciliationIT	cluster-wide DB↔ES counts read immediately after reindex Success; engine refresh lags → es < db	converge-poll until counts reconcile (keeps the cluster-wide umbrella; a real regression still never converges)
DistributedAutoTuneReindexUIIT	fuzzy q=<prefix> count also matched other parallel ui-it tests' entities (6900 vs 1500)	strict name.keyword prefix count, cluster-alias-aware index resolution, converge-poll, removed Thread.sleep
SimpleReindexTriggerUIIT, SelectiveFieldReindexUIIT, LongCompoundNameSearchUIIT, PipelineOwnerIndexUIIT	Awaitility .ignoreNoExceptions() aborted the poll on a transient Playwright/search error	.ignoreExceptions() so transient errors retry on the next tick
SelectiveFieldReindexUIIT	depended on the heavy global /data-quality page rendering under load	assert testCase/testSuite via their search index directly (the contract the DQ list is backed by)

SearchAvailableDuringReindexUIIT / SearchAvailableAllKindsDuringReindexUIIT are intentionally left unchanged — their .ignoreNoExceptions() is a deliberate fail-fast for the zero-downtime guarantee (a mid-flight blackout/duplicate must fail immediately; transient 503 shard-lag is already absorbed inside probeIndexToleratingShardLag).

Type of change:

• Bug fix (test reliability)

Tests:

Each modified test verified green on local testcontainers (Postgres + Elasticsearch, real Playwright browser):

• DbToEsCountReconciliationIT — 2/2 runs
• DistributedAutoTuneReindexUIIT, SimpleReindexTriggerUIIT, SelectiveFieldReindexUIIT, LongCompoundNameSearchUIIT, PipelineOwnerIndexUIIT — green

mvn test-compile and mvn spotless:check both pass.

🤖 Generated with Claude Code

Greptile Summary

This PR de-flakes the nightly search-it and ui-it reindex test suite by addressing root causes (ES refresh lag, fuzzy query scope bleed, and ignoreNoExceptions() aborting on transient errors) without weakening any assertions — two are in fact made stricter.

• DB↔ES reconciliation (DbToEsCountReconciliationIT): converts the immediate post-reindex count comparison into an Awaitility convergence poll so ES refresh lag no longer causes spurious failures; a real regression still never converges and fails.
• Distributed reindex count (DistributedAutoTuneReindexUIIT): replaces fuzzy q=<prefix> probe + Thread.sleep with a strict name.keyword prefix count via SearchAssertions and cluster-alias-aware IndexAliasInspector, removing the cross-test count bleed (6900 vs expected 1500) that was the primary flake source.
• ignoreNoExceptions() → ignoreExceptions() (SimpleReindexTriggerUIIT, LongCompoundNameSearchUIIT, PipelineOwnerIndexUIIT, SelectiveFieldReindexUIIT): Awaitility's ignoreNoExceptions() aborts the poll on any exception including transient Playwright/search errors; ignoreExceptions() lets those retry.
• DQ page dependency (SelectiveFieldReindexUIIT): replaces heavy /data-quality global page renders for testCase/testSuite with direct name.keyword index counts — the exact contract the DQ list is backed by.

Confidence Score: 5/5

All changes are test-only; no product behaviour is altered and no assertion is weakened.

Every change targets a clearly identified root cause — refresh lag, cross-test query bleed, or an incorrect Awaitility exception policy. The convergence poll in DbToEsCountReconciliationIT still catches genuine indexer regressions (they never converge). The switch from fuzzy q= to strict name.keyword prefix in DistributedAutoTuneReindexUIIT is strictly more precise. No logic shared with production code is touched.

No files require special attention.

Important Files Changed

Filename	Overview
openmetadata-integration-tests/src/test/java/org/openmetadata/it/tests/search/DbToEsCountReconciliationIT.java	Wraps the DB↔ES count reconciliation assertion in an Awaitility convergence poll to absorb post-reindex ES refresh lag; extracts collectCountMismatches() helper. Logic is sound.
openmetadata-integration-tests/src/test/java/org/openmetadata/playwright/scenarios/search/reindex/DistributedAutoTuneReindexUIIT.java	Replaces fuzzy q= probe + Thread.sleep + hardcoded alias map with strict name.keyword prefix count via SearchAssertions + IndexAliasInspector and Awaitility poll. Root cause of the 6900 vs 1500 flake is addressed cleanly.
openmetadata-integration-tests/src/test/java/org/openmetadata/playwright/scenarios/search/reindex/SelectiveFieldReindexUIIT.java	Replaces DataQualityListPage-based UI assertions for testCase/testSuite with direct search-index count checks; removes flaky heavy-page dependency. ignoreNoExceptions() → ignoreExceptions() for the shared pollUiAssertion.
openmetadata-integration-tests/src/test/java/org/openmetadata/playwright/scenarios/search/issues/LongCompoundNameSearchUIIT.java	Adds awaitDiscoverableInExplore() polling loop with short per-tick Playwright timeout; changes ignoreNoExceptions() → ignoreExceptions() so transient errors retry.
openmetadata-integration-tests/src/test/java/org/openmetadata/playwright/scenarios/search/reindex/SimpleReindexTriggerUIIT.java	Two ignoreNoExceptions() → ignoreExceptions() changes in separate Awaitility blocks. No logic changes.
openmetadata-integration-tests/src/test/java/org/openmetadata/playwright/scenarios/search/issues/PipelineOwnerIndexUIIT.java	Single-line change: ignoreNoExceptions() → ignoreExceptions() so transient Playwright/search errors retry instead of aborting the poll.

_{Reviews (1): Last reviewed commit: "test(search): de-flake nightly search-it..." | Re-trigger Greptile}

[github-actions]

❌ PR checklist incomplete

This PR cannot be merged until the following are addressed on its linked issue:

• Linked issue Flaky nightly search-it & ui-it reindex tests (timing/scope races, not product bugs) #29187 is not added to the Shipping project.

The fields live on the linked issue in the Shipping project (open the issue → right sidebar → Projects). After you set them, re-run this check (or push a commit) — issue/project changes do not re-trigger it automatically.

Maintainers can bypass this check by adding the skip-pr-checks label.

[gitar-bot]

Code Review ✅ Approved

Replaces flaky polling and hardcoded sleeps with robust converge-polling and strict keyword matching in search integration tests. No issues found.

Options

Display: compact → Showing less information.

Comment with these commands to change:

Compact

gitar display:verbose

_{Was this helpful? React with 👍 / 👎 | Gitar}