have you ever run disaster recovery (http://docs.ceph.com/docs/luminous/cephfs/disaster-recovery/). Try following steps stop mds.a and run following commands step by step cephfs-table-tool 0 reset session cephfs-journal-tool event recover_dentries summary cephfs-data-scan scan_links restart mds, don't mount any client, do a full filesystem scrub ceph daemon mds.a scrub_path / recursive repair On Sat, Aug 11, 2018 at 10:19 PM morfair@xxxxxxxxx <morfair@xxxxxxxxx> wrote: > > Hello all! > > My Ceph MDS is crashed and is no longer starting. > > How do I think the problem in this: > > 2018-08-10 16:59:18.147612 7f8d50037700 0 mds.0.cache.dir(0x604) > _fetched badness: got (but i already had) [inode 0x10001b99b20 [2,head] > ~mds0/stray1/10001b99b20 auth v100647297 s=540 n(v0 b540 1=1+0) > (iversion lock) cr={442500089=0-4194304@1} 0x558108ba1500] mode 33152 > mtime 2018-07-26 10:53:13.931856 > 2018-08-10 16:59:18.148084 7f8d50037700 -1 log_channel(cluster) log > [ERR] : loaded dup inode 0x10001b99b20 [2,head] v100727437 at > ~mds0/stray4/10001b99b20, but inode 0x10001b99b20.head v100647297 > already exists at ~mds0/stray1/10001b99b20 > > How to remove specific files with invalid inode? Or how to start my MDS > before monday work day? > > Full log: https://drive.google.com/open?id=16YMMht7X1M_Cu3jZh_gXWja3fqflVDNZ > > _______________________________________________ > ceph-users mailing list > ceph-users@xxxxxxxxxxxxxx > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |