Hi all,
We've got a containerized test cluster with 3 OSDs and ~ 220GiB of data. Shortly after upgrading from nautilus -> octopus, 2 of the 3 OSDs have started flapping. I've also got alarms about the MDS being damaged, which we've seen elsewhere and have a recovery process for, but I'm unable to run this (I suspect because I've only got 1 functioning OSD). My RGWs are also failing to start, again I suspect because of the bad state of OSDs. I've tried restarting all OSDs, rebooting all servers, checked auth (all looks fine) - but I'm still in the same state.
My OSDs seem to be failing at the "_open_alloc opening allocation metadata" step; looking at logs for each OSD restart, the OSD writes this log, then no logs for a few minutes and then logs:
bluestore(/var/lib/ceph/osd/ceph-1) _open_alloc loaded 132 GiB in 2930776 extents available 113 GiB
rocksdb: [db/db_impl.cc:390] Shutdown: canceling all background work
After that we seem to try starting up in a slightly different state and get a different set of errors:
bluefs _allocate failed to allocate 0x100716 on bdev 1, free 0xd0000; fallback to bdev 2
bluefs _allocate unable to allocate 0x100716 on bdev 2, free 0xffffffffffffffff; fallback to slow device expander
and eventually crash and log a heap of stack dumps.
I don't know what extents are but I seem to have a lot of them, and more than I've got capacity for? Maybe I'm running out of RAM or disk space somewhere, but I've got 21GB of free RAM on the server, and each OSD has a 350GiB device attached to it.
I'm wondering if anyone has seen anything like this before or can suggest next debug steps to take?
Cheers,
Dave
Full OSD logs surrounding the "_open_alloc opening allocation metadata" step:
Jul 23 00:07:13 condor_sc0 container_name/ceph-osd-1[1709]: 2021-07-23T00:07:13.818+0000 7f3de111bf40 4 rocksdb: EVENT_LOG_v1 {"time_micros": 1626998833819439, "job": 1, "event": "recovery_started", "log_files": [392088, 392132]}
Jul 23 00:07:13 condor_sc0 container_name/ceph-osd-1[1709]: 2021-07-23T00:07:13.818+0000 7f3de111bf40 4 rocksdb: [db/db_impl_open.cc:583] Recovering log #392088 mode 0
Jul 23 00:07:17 condor_sc0 container_name/ceph-osd-1[1709]: 2021-07-23T00:07:17.240+0000 7f3de111bf40 4 rocksdb: [db/db_impl_open.cc:583] Recovering log #392132 mode 0
Jul 23 00:07:17 condor_sc0 container_name/ceph-osd-1[1709]: 2021-07-23T00:07:17.486+0000 7f3de111bf40 4 rocksdb: EVENT_LOG_v1 {"time_micros": 1626998837486404, "job": 1, "event": "recovery_finished"}
Jul 23 00:07:17 condor_sc0 container_name/ceph-osd-1[1709]: 2021-07-23T00:07:17.486+0000 7f3de111bf40 1 bluestore(/var/lib/ceph/osd/ceph-1) _open_db opened rocksdb path db options compression=kNoCompression,max_write_buffer_number=4,min_write_buffer_number_to_merge=1,recycle_log_file_num=4,write_buffer_size=268435456,writable_file_max_buffer_size=0,compaction_readahead_size=2097152,max_background_compactions=2
Jul 23 00:07:17 condor_sc0 container_name/ceph-osd-1[1709]: 2021-07-23T00:07:17.524+0000 7f3de111bf40 1 freelist init
Jul 23 00:07:17 condor_sc0 container_name/ceph-osd-1[1709]: 2021-07-23T00:07:17.524+0000 7f3de111bf40 1 freelist _init_from_label
Jul 23 00:07:17 condor_sc0 container_name/ceph-osd-1[1709]: 2021-07-23T00:07:17.529+0000 7f3de111bf40 1 bluestore(/var/lib/ceph/osd/ceph-1) _open_alloc opening allocation metadata
Jul 23 00:07:18 condor_sc0 container_name/ceph-osd-1[1709]: 2021-07-23T00:07:18.238+0000 7f3de111bf40 1 HybridAllocator _spillover_range constructing fallback allocator
Jul 23 00:07:20 condor_sc0 container_name/ceph-osd-1[1709]: 2021-07-23T00:07:20.563+0000 7f3de111bf40 1 bluestore(/var/lib/ceph/osd/ceph-1) _open_alloc loaded 132 GiB in 2930776 extents available 113 GiB
Jul 23 00:07:20 condor_sc0 container_name/ceph-osd-1[1709]: 2021-07-23T00:07:20.563+0000 7f3de111bf40 4 rocksdb: [db/db_impl.cc:390] Shutdown: canceling all background work
Jul 23 00:07:20 condor_sc0 container_name/ceph-osd-1[1709]: 2021-07-23T00:07:20.565+0000 7f3de111bf40 4 rocksdb: [db/db_impl.cc:563] Shutdown complete
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
