zhangshenghang commented on PR #11562:
URL: https://github.com/apache/seatunnel/pull/11562#issuecomment-5862863796

   Thanks @DanielLeens — both blockers are addressed in f8abb17f5f (on top of a 
fresh `upstream/dev` merge).
   
   - **Issue 1 (unbounded HTTP call defeating the deadline)**: `HttpUtil`'s 
control client — the one used for `get_load_state` polling and the commit/abort 
control requests — now carries a `RequestConfig` with 30s connect and socket 
timeouts. The values match the connector's own source-side defaults 
(`DorisSourceOptions`: 30s/30s) and sit well below the default 300s 
`sink.visibility-timeout-ms`, so a single hung request is retried and the outer 
deadline stays meaningful. The stream-load *upload* client deliberately keeps 
no socket timeout so large body transfers are not cut off mid-flight.
   - **Issue 2 (docs)**: `sink.visibility-timeout-ms` is now in the sink 
options table of `docs/en/connectors/sink/Doris.md` and `docs/zh/...`, with a 
short "Data Visibility Wait (2PC)" section in both languages explaining the 
COMMITTED-before-VISIBLE window, the `get_load_state` polling, 
ABORTED/CANCELLED as terminal failures, and that both the loop and the 
underlying control requests are bounded.
   
   Verified locally: `connector-doris` module tests 137/137 green (incl. 
`DorisCommitterTest`). Issue 3 (regression test for repeated `IOException` 
until deadline exhaustion) is still open as the agreed non-blocking follow-up — 
happy to add it in a follow-up commit if you want it in this PR.


-- 
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