Sailesh Mukil has posted comments on this change. Change subject: IMPALA-5199: prevent hang on empty row batch exchange ......................................................................
Patch Set 2: The fix looks good to me, and should work fine for most cases. One case I'm worried about is that, under heavy load, if a receiver deregisters itself early (due to a LIMIT, etc.), and the sender sends the 'eos' after STREAM_EXPIRATION_TIME_MS, the query would have previously succeeded, even though the sender couldn't find the receiver in CloseSender(). And the query would be right to succeed in this case since the 'eos' here is spurious as the receiver is already gone. Although, I know this counts as succeeding "by mistake". Now, we would be failing these queries. And given that we see the 'DATASTREAM_SENDER_TIMEOUT' fairly often, that would regress a few workloads. What do you think? -- To view, visit http://gerrit.cloudera.org:8080/8005 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: Ib370ebe44e3bb34d3f0fb9f05aa6386eb91c8645 Gerrit-PatchSet: 2 Gerrit-Project: Impala-ASF Gerrit-Branch: master Gerrit-Owner: Tim Armstrong <tarmstr...@cloudera.com> Gerrit-Reviewer: Sailesh Mukil <sail...@cloudera.com> Gerrit-HasComments: No