Todd Lipcon has posted comments on this change. Change subject: log: Mark allocation finished even if allocation had an error ......................................................................
Patch Set 2: (1 comment) http://gerrit.cloudera.org:8080/#/c/3234/2/src/kudu/consensus/log.cc File src/kudu/consensus/log.cc: Line 188: // TODO If a single transaction fails to append, should we hm, this TODO is now scaring me. If we fail to append one operation (eg because the disk is full, or because of a temporary IO error) we could keep going and have a "gap" in txids? Have you checked all call sites to make sure we FATAL on a failure to append to the log? (I see that this was already the case in RELEASE builds, so you aren't making anything worse than it was, but it is uncovering a scariness here) -- To view, visit http://gerrit.cloudera.org:8080/3234 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: If22bf946a42d0ec32c35164acd9e6e6cef18dcc3 Gerrit-PatchSet: 2 Gerrit-Project: kudu Gerrit-Branch: master Gerrit-Owner: Mike Percy <mpe...@apache.org> Gerrit-Reviewer: Adar Dembo <a...@cloudera.com> Gerrit-Reviewer: Kudu Jenkins Gerrit-Reviewer: Mike Percy <mpe...@apache.org> Gerrit-Reviewer: Todd Lipcon <t...@apache.org> Gerrit-HasComments: Yes