Adar Dembo has posted comments on this change. Change subject: KUDU-1328: crash when remote bootstrap session races with tablet deletion ......................................................................
Patch Set 2: (1 comment) http://gerrit.cloudera.org:8080/#/c/2193/2/src/kudu/tserver/remote_bootstrap_session.cc File src/kudu/tserver/remote_bootstrap_session.cc: Line 105: RETURN_NOT_OK(tablet_peer_->log()->GetLogReader()->GetSegmentsSnapshot(&log_segments_)); > ah, it was a crash log that JD sent me from the same cluster where we saw t As we discussed offline, this is still a very real and live issue. It's also trickier to handle than the other race because LogReader isn't supposed to be shared. The work done to prevent leaking FDs has also (I think) broken the contract outlined in log.h: // Returns a reader that is able to read through the previous // segments. The reader pointer is guaranteed to be live as long // as the log itself is initialized and live. LogReader* GetLogReader() const; Will have to think about this some more. -- To view, visit http://gerrit.cloudera.org:8080/2193 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: I6a487ba4f37383479170519866509f3f3d88dc5d Gerrit-PatchSet: 2 Gerrit-Project: kudu Gerrit-Branch: master Gerrit-Owner: Adar Dembo <[email protected]> Gerrit-Reviewer: Adar Dembo <[email protected]> Gerrit-Reviewer: Jean-Daniel Cryans Gerrit-Reviewer: Kudu Jenkins Gerrit-Reviewer: Todd Lipcon <[email protected]> Gerrit-HasComments: Yes
