apt-get install multipath-tools
Edited /etc/multipath.conf
:
defaults { polling_interval 30 failback immediate no_path_retry 5 rr_min_io 100 path_checker tur user_friendly_names yes } multipaths { multipath { wwid 360050768019101ca2800000000000031 alias IBM-2145 } }
multipath -l now gives:
IBM-2145 (360050768019101ca2800000000000031) dm-4 IBM,2145
[size=2.0T][features=1 queue_if_no_path][hwhandler=0]
\_ round-robin 0 [prio=0][enabled]
\_ 3:0:0:0 sdb 8:16 [active][undef]
\_ round-robin 0 [prio=0][enabled]
\_ 3:0:1:0 sdc 8:32 [active][undef]
Now, /dev/sdc
, which we used to mount, is in use by multipathd
, and we cannot mount it any more.
/etc/init.d/multipath-tools restart
mount -t xfs /dev/disk/by-id/dm-uuid-mpath-360050768019101ca2800000000000031 /srv/linux/ runs forever with no CPU load, and cannot be killed, so I think it is waiting for I/O.
/bin/sh -f /usr/sbin/xfs_check /dev/disk/by-uuid/../../mapper/IBM-2145 also ends up waiting for I/O.
Furthermore, /var/log/daemon.log
reads:
...
Feb 23 18:06:16 si17 multipathd: 8:16: mark as failed
Feb 23 18:06:17 si17 multipathd: 8:32: mark as failed
Feb 23 18:06:46 si17 multipathd: 8:16: reinstated
Feb 23 18:06:46 si17 multipathd: 8:16: mark as failed
Feb 23 18:06:47 si17 multipathd: 8:32: reinstated
Feb 23 18:06:47 si17 multipathd: 8:32: mark as failed
Feb 23 18:07:16 si17 multipathd: 8:16: reinstated
Feb 23 18:07:16 si17 multipathd: 8:16: mark as failed
Feb 23 18:07:17 si17 multipathd: 8:32: reinstated
Feb 23 18:07:17 si17 multipathd: 8:32: mark as failed
Feb 23 18:07:46 si17 multipathd: 8:16: reinstated
Feb 23 18:07:46 si17 multipathd: 8:16: mark as failed
Feb 23 18:07:47 si17 multipathd: 8:32: reinstated
Feb 23 18:07:47 si17 multipathd: 8:32: mark as failed
Feb 23 18:08:16 si17 multipathd: 8:16: reinstated
Feb 23 18:08:16 si17 multipathd: 8:16: mark as failed
Feb 23 18:08:17 si17 multipathd: 8:32: reinstated
Feb 23 18:08:17 si17 multipathd: 8:32: mark as failed
Feb 23 18:08:46 si17 multipathd: 8:16: reinstated
Feb 23 18:08:46 si17 multipathd: 8:16: mark as failed
Feb 23 18:08:47 si17 multipathd: 8:32: reinstated
Feb 23 18:08:47 si17 multipathd: 8:32: mark as failed
And multipath -l now gives:
IBM-2145 (360050768019101ca2800000000000031) dm-4 IBM,2145
[size=2.0T][features=1 queue_if_no_path][hwhandler=0]
\_ round-robin 0 [prio=0][enabled]
\_ 3:0:0:0 sdb 8:16 [failed][undef]
\_ round-robin 0 [prio=0][enabled]
\_ 3:0:1:0 sdc 8:32 [failed][undef]
(It seems to cycle through “failed” and “active”.)
Also, after /etc/init.d/multipath-tools stop, I/O still fails on both /dev/sdc
and /dev/sdb
.
And when the multipath-tools
are running, I/O fails on /dev/mapper/IBM-2145
.
Rebooting solves. We can now at least mount /dev/mapper/IBM-2145 read-only. Still it is a Windows solution. And I don't understand what went wrong.
And now the problem is back: the filesystem on /dev/mapper/IBM-2145
is mounted read-only (and without filesystem cehck) on /srv/linux
, but both a find on the mount point and an xfs_check on the block device end up waiting for I/O.
And this time, rebooting ends with a machine that responds to ping
, but nothing else.
I have no idea whether it hangs in shutdown (the unmount?) or didn't come up properly.
Note | |
---|---|
Arjan solved this one. It turns out that the add-in card was broken. He replaced it with a new one, and in the process replaced the entire server too. |
Note | |
---|---|
This was something of a bad start, but things got better. The rest of this adventure is in http://www.cs.rug.nl/~jurjen/iwi-howtos/linux/Osis. |