ryanmeowy commented on PR #12711:
URL: https://github.com/apache/seatunnel/pull/12711#issuecomment-6097476722

   CI is green on this PR: `Build` pass, both `elasticsearch-connector-it` legs 
(`11` and `17`) successful — [run 
38043755930](https://github.com/ryanmeowy/seatunnel/actions/runs/38043755930).
   
   Baseline this fixes, for reviewers: on the current `dev` base (`0278a6a74`) 
the same job fails with `Tests run: 94, Failures: 7 ... in ElasticsearchIT` 
(module total: 116 tests, 7 failures) — [run 
37927712604](https://github.com/apache/seatunnel/actions/runs/37927712604). All 
seven are the three slice tests:
   
   - `ElasticsearchIT.testElasticsearchWithPITSlice` — assertion at 
`ElasticsearchIT.java:495`
   - `testElasticsearchWithPITSliceQueued` — `:507`
   - `testElasticsearchWithScrollSlice` — `:518`
   
   Failure breakdown: `org.opentest4j.AssertionFailedError: iterable lengths 
differ, expected: <11> but was: <0>` (4x), `expected: <11> but was: <10>` (1x), 
and on the two Spark variants `iterable contents differ at index [1]` (rows 
shifted, e.g. expected `c_int: 11`, got `16`). These tests read the sink index 
once and assert immediately, while the Elasticsearch sink writes through 
`esRestClient.bulk(...)` with no refresh policy, so the documents are not 
guaranteed to be visible yet. In the same dev run the outcome differs per 
engine variant — flink-1.13.6 and flink-1.18.0 passed where flink-1.15.3 and 
flink-1.20.1 failed, and the Zeta variants never failed — so it is 
timing-dependent rather than a slicing logic bug.
   
   #12642 (`9b7f1c863`) switched the other tests in this file to `Awaitility`; 
these three were the remaining call sites without it. This PR wraps their 
assertions in the same 
`Awaitility.await().atMost(INDEX_REFRESH_TIMEOUT_SECONDS, ...)` pattern the 
neighbouring tests already use, so the read is retried for up to 60s instead of 
asserted once.
   
   @zhangshenghang is applying the same change to `ElasticsearchAuthIT` in 
#12643. Either folding this into that PR or keeping it separate works for me.
   
   This is also the only red check on #12708; once this lands I will rebase 
#12708 and re-run it.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to