Thanks a ton! Yes, restart mds fixed this. But can’t confirm it hit bug 50840, seems when we read huge small files will hit this! (means more than 10,000 small files in one directory ). Thanks Xiong > 2022年8月26日 19:13，Stefan Kooman <stefan@xxxxxx> 写道： > > On 8/26/22 12:33, zxcs wrote: >> Hi, experts >> we have a cephfs cluster with 15.2.* version and kernel mount, today there is a health report mds slow request as below, i checked this mds log, seems it report some slow request for a long time. >> mds report: >> 1 MDSs report slow requests >> mds log: >> log_channel(cluster) log [WRN] : slow request 34616.878139 seconds old, received at 2022-08-26T08:49:16.400430+0800: client_request(client.100807545:2601765 getattr >> i know we can restart mds to fix this(may be), but seems there only one directory hang, called A, (means when i ls -lrth /ceph/path/A, it stuck), and list other directory no issue. > Might be this bug: https://tracker.ceph.com/issues/50840 > We hit this bug. A restart of the MDS is necessary. What version of Octopus do you run? This is fixed in Octopus 15.2.17 [1,2]. > > So if you hit this bug (which might be difficult to tell) you can update the MDS and restart the MDS with the new version. > >> my question is how can we fix this without remount ceph on this node or restart mds (this will impact other uses). > > Gr. Stefan > > [1]: https://docs.ceph.com/en/latest/releases/octopus/#changelog > [2]: https://tracker.ceph.com/issues/51202 > _______________________________________________ > ceph-users mailing list -- ceph-users@xxxxxxx > To unsubscribe send an email to ceph-users-leave@xxxxxxx _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |