Matthew Jacobs has posted comments on this change. Change subject: IMPALA-4411: Kudu inserts violate lock ordering and could deadlock ......................................................................
Patch Set 1: (3 comments) http://gerrit.cloudera.org:8080/#/c/4895/1//COMMIT_MSG Commit Message: PS1, Line 17: Running an exhaustive test job, and stress tests : manually. > Did you reproduce the hang? It should be easy to do if you add a sleep call I wasn't able to reproduce it easily. I tried inserting sleeps and running stress and cancellation tests, but no luck. I think it may have required a rare combination of events across a number of threads, e.g. like the one mentioned by Tim in his related patch for fixing another case of mis-ordering the same locks. http://gerrit.cloudera.org:8080/#/c/4895/1/be/src/service/query-exec-state.cc File be/src/service/query-exec-state.cc: PS1, Line 532: // This should be safe to access on coord_ after Wait() has been called. > This is a bit vague. Is it safe or not? Done PS1, Line 535: "Updating session latest observed Kudu timestamp: " << latest_kudu_ts > Add the session ID here, otherwise this message is not going to be useful ( Done -- To view, visit http://gerrit.cloudera.org:8080/4895 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: I2aa36fce78525a80a6b880e1b668105006bc1425 Gerrit-PatchSet: 1 Gerrit-Project: Impala-ASF Gerrit-Branch: master Gerrit-Owner: Matthew Jacobs <[email protected]> Gerrit-Reviewer: Henry Robinson <[email protected]> Gerrit-Reviewer: Matthew Jacobs <[email protected]> Gerrit-Reviewer: Sailesh Mukil <[email protected]> Gerrit-Reviewer: Tim Armstrong <[email protected]> Gerrit-HasComments: Yes
