[Sheepdog] qemu-system-x86_64 does not work after expanding the cluster

[MORITA Kazutaka]

At Tue, 16 Aug 2011 00:12:15 +0200,
Valerio Pachera wrote:
> 
> Hi have 2 servers with the same partirion schema:
> sdb6 xfs /mnt/sheepdog 20G
> 
> I created a disk called test
> 
> # collie vdi list
>   name        id    size    used  shared    creation time   vdi id
> ------------------------------------------------------------------
>   test         1   10 GB  536 MB  0.0 MB 2011-08-14 15:28   7c2b25
> 
> I run a vm using a systemrescuecd iso:
> qemu-system-x86_64 sheepdog:test -cdrom
> /root/systemrescuecd-x86-2.3.0.iso -boot order=d -m 1024 -smp 1 -vnc
> :1 -daemonize -no-kvm
> 
> Unfortunatrly these two servers do not have vt-tecnology so I used
> qemu-kvm vithout it (but this doesn't matter for us).
> From system rescue I run
> dd if=/dev/zero of=/dev/sda bs=1M count=512
> 
> Happy to see it was working, I shut off the vm.
> I created a new partition only on server1 of the same size of the
> other, but on another phisical disk:
> sdc6 xfs /mnt/sheepdog2 20G
> sheep /mnt/sheepdog2
> 
> I was windering when does sheepdog start to balance the cluser. I
> aspect to see data be moved from /mnt/sheepdog right away or when a vm
> try to write something new.
> Right away nothing happened. So I run again 'qemu-system-x86_64
> sheepdog:test ...'  but, this time, the command doesn't go in
> background and simply do nothing!
> 
> I also tried to create sdc6 on server2 and run 'sheep /mnt/sheepdog2',
> but qemu is still blocked when I run it.
> No strange things in dmesg.
> 
> On server2:
> more /mnt/sheepdog2/sheep.log
> Aug 15 11:59:48 jrnl_recover(2221) Openning the directory
> /mnt/sheepdog2//journal/00000000/.
> Aug 15 11:59:48 create_listen_ports(163) can't create a listen fd

You need to specify another port number to the second daemon:

  $ sheep /mnt/sheepdog2 -p 7001


> 
> On server1 the same file is empty.
> 
> This is the last part of '/mnt/sheepdog/sheep.log' on server2.
> Aug 14 02:33:05 worker_routine(208) started this thread 12
> Aug 14 02:33:05 worker_routine(208) started this thread 14
> Aug 14 02:33:05 set_addr(1595) addr = 192.168.100.1, port = 7000
> Aug 14 02:33:05 main(144) Sheepdog daemon (version 0.2.3) started
> Aug 14 02:33:05 worker_routine(208) started this thread 16
> Aug 14 02:33:05 worker_routine(208) started this thread 17
> Aug 14 02:33:05 sd_confchg(1493) confchg nodeid 164a8c0
> Aug 14 02:33:05 sd_confchg(1495) 2 0 1
> 
> And this on server1
> Aug 14 02:32:38 worker_routine(208) started this thread 30
> Aug 14 02:32:38 worker_routine(208) started this thread 31
> Aug 14 02:32:38 set_addr(1595) addr = 192.168.100.2, port = 7000
> Aug 14 02:32:38 worker_routine(208) started this thread 32
> Aug 14 02:32:38 worker_routine(208) started this thread 33
> Aug 14 02:32:38 main(144) Sheepdog daemon (version 0.2.3) started
> Aug 14 02:32:38 worker_routine(208) started this thread 34
> Aug 14 02:32:38 sd_confchg(1493) confchg nodeid 264a8c0
> Aug 14 02:32:38 sd_confchg(1495) 1 0 1
> Aug 14 02:32:38 sd_confchg(1499) [0] node_id: 40151232, pid: 27495,
> reason: 180493992
> Aug 14 02:32:38 sd_confchg(1513) allow new confchg, 0x46456f0
> 
> ----
> Debian squeeze, 2.6.32-5-amd64, sheepdog 0.2.3, qemu-kvm 0.15.0.

Sorry, I couldn't reproduce the problem.  Are you using the
development source available from the git repoisitory?  If no, can you
try it?


Thanks,

Kazutaka