[sheepdog-users] vdi problem

[Valerio Pachera]

Yesterday I cleared sheep.log on my 3 hosts (production cluster).
The last night the guest 'backup' gave the same problem when it stated
receiving backup data.

This morning I see

root at sheepdog002:~# cat /mnt/ST2000DM001-1CH164_W1E2N5GM/sheep.log
Jun 12 16:50:56 [gway 12895] wait_forward_request(176) poll timeout 1,
disks of some nodes or network is busy. Going to poll-wait again
Jun 12 16:51:33 [gway 12904] wait_forward_request(176) poll timeout 1,
disks of some nodes or network is busy. Going to poll-wait again
Jun 12 16:51:35 [gway 12909] wait_forward_request(176) poll timeout 1,
disks of some nodes or network is busy. Going to poll-wait again

sheepdog001 and sheepdog002 are empty.

>From /var/log/messages of sheepdog002 I see

  Jun 12 19:22:29 sheepdog002 kernel: [197497.557908] Pid: 0, comm:
swapper/1 Tainted: G           O 3.2.0-4-amd64 #1 Debian 3.2.41-2
  Jun 12 19:22:29 sheepdog002 kernel: [197497.557915] Call Trace:
  <cut>
  Jun 12 19:22:31 sheepdog002 kernel: [197500.089950] Pid: 14513,
comm: sheep Tainted: G           O 3.2.0-4-amd64 #1 Debian 3.2.41-2
  Jun 12 19:22:31 sheepdog002 kernel: [197500.089957] Call Trace:

And it repeats several times

root at sheepdog002:~# free -m
             total       used       free     shared    buffers     cached
Mem:          1884       1785         98          0          0        699
-/+ buffers/cache:       1085        799
Swap:         7632       2719       4913

My considerations:

it seems like a lack of memory. I can't figure out the cause.
It sounds strange because the guest may use 1G maximum of ram, so the
host has 1G for it self (xorg is not even installed, so no much ram
used by debian itself).
The guest is anyway running only nfs-kernel-server, so it should not
use lot's of ram, even with high traffic.
I noticed that, writing data inside the guest by dd,  doesn't trigger
the 'DRDY error' but write speed is low (7M/s for 512Mbye instead of
80-90M/s).
This doesn't happen on a host with more ram (cpu is also an i5 instead
of an amd turion).
Sheep daemon should not use lot's of memory right?
Right now the guest died without any reason/interaction
  Jun 13 11:19:23 sheepdog002 kernel: [254912.009837]
qemu-system-x86[12290] general protection ip:7f0c10afc386
sp:7f0bcd5ec370 error:0 in libz.so.1.2.7[7f0c10af0000+16000
All hosts are running QEMU emulator version 1.4.90.

I'm going to update qemu.

What do you think about that?