Spurious 'incorrect nilfs2 checksum' breaking ceph OSD

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi,

This has bitten us a couple of times now (such that we're considering
re-building util-linux with the nilfs2 code commented out), so I'm
wondering if anyone else has seen it [and noting the failure mode in
case anyone else is confused in future]

We see this with our setup of rotating media for the osd, NVMe partition
for journal.

What happens is that sometimes an osd refuses to start up, complaining
that /var/lib/ceph/osd/ceph-XXX/journal is missing.

inspecting that file will show it's a broken symlink to an entry in
/dev/disk/by-partuuid:

/var/lib/ceph/osd/ceph-388/journal: broken symbolic link to
/dev/disk/by-partuuid/d2ace848-7e2d-4395-a195-a4428631b333

If you inspect the relevant partition, you see that it has the matching
block id:

blkid /dev/nvme0n1p11
/dev/nvme0n1p11: PARTLABEL="ceph journal"
PARTUUID="d2ace848-7e2d-4395-a195-a4428631b333

And, if you look in syslog, you'll see this:

Jan  4 09:25:29 sto-3-3 systemd-udevd[107317]: incorrect nilfs2 checksum
on /dev/nvme0n1p11

The problem is that the nilfs2 checker is too promiscuous, looking for a
relatively short magic number (0x3434) in 2 different places (location
0x400, and (((part_size-512)/8)*512)). So sometimes you'll be unlucky
and have a ceph journal that matches, at which point the nilfs2 prober
find an invalid checksum, and so systemd/udevd doesn't create the
/dev/disk/by-partuuid link.

You can work around this by making the symlink by hand when the failure
occurs; I also understand that the nilfs2 prober in util_linux 2.29 is
more robust (but that's not in any LTS distributions yet, so I've not
tested it).

Regards,

Matthew

util-linux issue: https://github.com/karelzak/util-linux/issues/361
Ubuntu bug:
https://bugs.launchpad.net/ubuntu/+source/util-linux/+bug/1653936


-- 
 The Wellcome Trust Sanger Institute is operated by Genome Research 
 Limited, a charity registered in England with number 1021457 and a 
 company registered in England with number 2742969, whose registered 
 office is 215 Euston Road, London, NW1 2BE. 
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com



[Index of Archives]     [Information on CEPH]     [Linux Filesystem Development]     [Ceph Development]     [Ceph Large]     [Linux USB Development]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [xfs]


  Powered by Linux