Adar Dembo has posted comments on this change. Change subject: log: shut down appender thread when idle ......................................................................
Patch Set 2: (1 comment) http://gerrit.cloudera.org:8080/#/c/6856/1/src/kudu/consensus/log.cc File src/kudu/consensus/log.cc: Line 256: &worker_state_, WORKER_STOPPED, WORKER_ACTIVE); > hard to repro in a test without injecting various sleeps. I did manually in I looked at mt-log-test and I don't think it's geared to trigger this, because all of the writers are running all at once, and then not at all. That is, there's no staggering, so there's no window of time where the queue will be empty of entries for an entire second. Maybe mt-log-test could reliably trigger this if you made kIdleThreshold configurable and turned it way down (say 1ms) for the test? That would probably be a good stress test in and of itself, since it'd trigger many active->stopped and stopped->active transitions. Broadly speaking I don't find MAYBE... style fault injection to be polluting, but I don't want to ask to you spend hours building a test that, in the best of times, rarely triggers the race. So, if it's easily reproducible by tweaking the timeout (or injecting some latency), I'd support it, but not if it necessitates a new test with hundreds of lines of code. -- To view, visit http://gerrit.cloudera.org:8080/6856 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: Id40a4cfcec96198b537c2f50be7ff1204faf96d2 Gerrit-PatchSet: 2 Gerrit-Project: kudu Gerrit-Branch: master Gerrit-Owner: Todd Lipcon <t...@apache.org> Gerrit-Reviewer: Adar Dembo <a...@cloudera.com> Gerrit-Reviewer: Kudu Jenkins Gerrit-Reviewer: Tidy Bot Gerrit-Reviewer: Todd Lipcon <t...@apache.org> Gerrit-HasComments: Yes
