On Mon, Aug 26, 2019 at 6:57 PM thoralf schulze <t.schu...@tu-berlin.de> wrote: > > hi Zheng, > > On 8/21/19 4:32 AM, Yan, Zheng wrote: > > Please enable debug mds (debug_mds=10), and try reproducing it again. > > please find the logs at > https://www.user.tu-berlin.de/thoralf.schulze/ceph-debug.tar.xz . > > we managed to reproduce the issue as a worst case scenario: before > snapshotting, juju-d0f708-5-lxd-1 and juju-d0f708-10-lxd-1 were the > active mds's and juju-d0f708-3-lxd-1 and juju-d0f708-9-lxd-1 standbys. > we created the snapshot at ~08:11:50, a little later the failover > happened and juju-d0f708-5-lxd-1 and juju-d0f708-10-lxd-1 went mia. a > little later still, the now-active juju-d0f708-3-lxd-1 and > juju-d0f708-9-lxd-1 mds's dropped out of the cluster as well. we started > to restart all mds daemons at ~08:16. > > thank you very much & with kind regards, > t. >
I tracked down the bug https://tracker.ceph.com/issues/41434 Thanks Yan, Zheng _______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
