[lustre-discuss] backup zfs MDT or migrate from ZFS back to ldiskfs
Stu Midgley
sdm900 at gmail.com
Thu Jul 20 21:54:15 PDT 2017
Afternoon
I have an MDS running on spinning media and wish to migrate it to SSD's.
Lustre 2.9.52
ZFS 0.7.0-rc3
How do I do it?
This is a loaded question :)
The MDT is using ~2TB of space. I used the zfs send | zfs receive method
to no avail. It was just too slow (I killed it after a week), gave no
indication of how far though it was and the destination FS was using ~4TB
of space (more than the 2TB of the source FS).
I have tried creating a snapshot, mounting it and rsyncing the contents to
the new FS. That fails with
mount.lustre: mount localData/t1-MDT0001 at /localData/t1-MDT0001 failed:
Invalid argument
This may have multiple causes.
Are the mount options correct?
Check the syslog for more info.
and from dmesg
[6982824.344872] Lustre: 401:0:(client.c:2113:ptlrpc_expire_one_request())
@@@ Request sent has timed out for slow reply: [sent 1500606249/real
1500606249] req at ffff880d70b56300 x1566221596712816/t0(0)
o38->t1-MDT0000-lwp-OST0000 at 0@lo:12/10 lens 520/544 e 0 to 1 dl 1500606260
ref 1 fl Rpc:XN/0/ffffffff rc 0/-1
[6982829.865323] LustreError: 18881:0:(obd_config.c:574:class_setup())
setup t1-MDT0000 failed (-22)
[6982829.868476] LustreError:
18881:0:(obd_config.c:1710:class_config_llog_handler()) MGC172.16.250.40 at tcp:
cfg command failed: rc = -22
[6982829.874691] Lustre: cmd=cf003 0:t1-MDT0000 1:t1-MDT0000_UUID 2:0
3:t1-MDT0000-mdtlov 4:f
[6982829.881030] LustreError: 15b-f: MGC172.16.250.40 at tcp: The
configuration from log 't1-MDT0000'failed from the MGS (-22). Make sure
this client and the MGS are running compatible versions of Lustre.
[6982829.887555] LustreError:
18811:0:(obd_mount_server.c:1352:server_start_targets()) failed to start
server t1-MDT0000: -22
[6982829.890670] LustreError:
18811:0:(obd_mount_server.c:1844:server_fill_super()) Unable to start
targets: -22
[6982829.893651] LustreError: 18811:0:(obd_config.c:621:class_cleanup())
Device 443 not setup
[6982830.300528] Lustre: server umount t1-MDT0000 complete
[6982830.303158] LustreError:
18811:0:(obd_mount.c:1496:lustre_fill_super()) Unable to mount (-22)
While doing the rsync -aXv I get weird error messages
file has vanished: "/mnt/mdt.s1/oi.7/0x200000007:0x1:0x0/mpartition"
which is on the ro snapshot of the original FS. It gets even weirder...
170721112705 pud40:mnt# ls -l
/mnt/mdt.s1/oi.7/0x200000007:0x1:0x0/mpartition
ls: cannot read symbolic link
/mnt/mdt.s1/oi.7/0x200000007:0x1:0x0/mpartition: No such file or directory
lrwxrwxrwx 1 root root 6 Mar 9 07:50
/mnt/mdt.s1/oi.7/0x200000007:0x1:0x0/mpartition
# strace -f -e readlink -- ls -l
/mnt/mdt.s1/oi.7/0x200000007:0x1:0x0/mpartition
readlink("/mnt/mdt.s1/oi.7/0x200000007:0x1:0x0/mpartition", 0xc8c320, 7) =
-1 ENOENT (No such file or directory)
ls: cannot read symbolic link
/mnt/mdt.s1/oi.7/0x200000007:0x1:0x0/mpartition: No such file or directory
lrwxrwxrwx 1 root root 6 Mar 9 07:50
/mnt/mdt.s1/oi.7/0x200000007:0x1:0x0/mpartition
+++ exited with 2 +++
# stat /mnt/mdt.s1/oi.7/0x200000007:0x1:0x0/mpartition
File: ‘/mnt/mdt.s1/oi.7/0x200000007:0x1:0x0/mpartition’stat: cannot read
symbolic link ‘/mnt/mdt.s1/oi.7/0x200000007:0x1:0x0/mpartition’: No such
file or directory
Size: 6 Blocks: 2 IO Block: 512 symbolic link
Device: 41h/65d Inode: 738 Links: 1
Access: (0777/lrwxrwxrwx) Uid: ( 0/ root) Gid: ( 0/ root)
Access: 2017-07-21 10:44:27.000000000 +0800
Modify: 2017-03-09 07:50:43.000000000 +0800
Change: 2017-07-21 10:44:27.000000000 +0800
Birth: -
so the file is in the directory listing as a link but not on the file
system??
So, this leads me to think that lustre is using features of ZFS that linux
can't access? or something??
It also raised the issue... how would I migrate off ZFS if I wanted to go
back to ldiskfs for the MDT?
Any ideas appreciated
Thanks.
--
Dr Stuart Midgley
sdm900 at sdm900.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lustre.org/pipermail/lustre-discuss-lustre.org/attachments/20170721/3954a2f8/attachment.htm>
More information about the lustre-discuss
mailing list