> On Sept. 7, 2017, 12:05 p.m., Alexander Rukletsov wrote: > > src/slave/constants.hpp > > Line 74 (original), 74 (patched) > > <https://reviews.apache.org/r/62140/diff/2/?file=1816999#file1816999line74> > > > > I know it is hard to justify specific values, but `1s` feels too > > aggressive.
>From the scenario described in MESOS-7939, this needs to be less than 10 secs >to prevent agent failure, or less than 5s for agent to recover from failure. >Also the check itself is pretty cheap, and that's why I'm using such an >aggresive value. - Chun-Hung ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/62140/#review184814 ----------------------------------------------------------- On Sept. 7, 2017, 4:01 a.m., Chun-Hung Hsiao wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/62140/ > ----------------------------------------------------------- > > (Updated Sept. 7, 2017, 4:01 a.m.) > > > Review request for mesos, Alexander Rukletsov, Benjamin Mahler, Jie Yu, and > Vinod Kone. > > > Bugs: MESOS-7939 > https://issues.apache.org/jira/browse/MESOS-7939 > > > Repository: mesos > > > Description > ------- > > Make the default value of `disk_watch_interval` to 1 second so GC can be > kicked in early enough before the agent fails due to a "No space left on > disk" failure. This change would also trigger the 1st GC earlier during > agent startup, so it can recover from the above scenario. > > > Diffs > ----- > > src/slave/constants.hpp 435c37e28987119c4ef43c9dad61ac052020acd8 > > > Diff: https://reviews.apache.org/r/62140/diff/2/ > > > Testing > ------- > > make test > > > Thanks, > > Chun-Hung Hsiao > >
