[Sheepdog] VDI's disappear in cluster

[Kazutaka MORITA]

Hi, Chris

On Thu, Feb 17, 2011 at 11:31 PM, Chris Hellkvist
<chris.hellkvist at googlemail.com> wrote:
> Hi,
>
> i have some problems with sheepdog; it looks to me like VDI's are
> getting lost in the sheepdog cluster:
> Sheepdog is running on 5 nodes, seems all fine:
>
> root at 2tb-incep-9:~# collie node list
>  Idx   Node id (FNV-1a) - Host:Port
> ------------------------------------------------
>  0     2703573cdc30b0ca - 10.250.241.117:7000
>  1     6967166567a2e289 - 10.250.241.118:7000
> * 2     a52f6bf8425ebf64 - 10.250.241.119:7000
>  3     bafd0bd834340c10 - 10.250.241.115:7000
>  4     ddd81841dcb2f16f - 10.250.241.116:7000
>
> I formatted the cluster:
> root at 2tb-incep-9:~# collie cluster format --copies=3
>
> And created a VDI with qemu:
> root at 2tb-incep-9:~# qemu-img create sheepdog:vditest 4G
> Formatting 'sheepdog:vditest', fmt=raw size=4294967296
> root at 2tb-incep-9:~# collie vdi list
>  name        id    size    used  shared    creation time   vdi id
> ------------------------------------------------------------------
>  vditest      1  4.0 GB  0.0 MB  0.0 MB 2011-02-17 15:23   c584fa
>
> So to this point all looks fine. Then i try to run qemu with this VDI,
> this fails:
> root at 2tb-incep-9:~# qemu-system-x86_64 -cdrom
> debian-6.0.0-amd64-netinst.iso -boot d -vnc :1 -net nic -net
> tap,ifname=tap0,script=no -daemonize sheepdog:vditest
> qemu-system-x86_64: -daemonize: cannot get vdi info, No vdi found, vditest 0
> qemu: could not open disk image sheepdog:vditest: Operation not permitted
>
> And now, the VDI in the clister is gone:
> root at 2tb-incep-9:~# collie vdi list
>  name        id    size    used  shared    creation time   vdi id
> ------------------------------------------------------------------
>
> Any ideas on that, or at least how to debug this? We are running 0.2.1
> from the git repo...

Sorry for the inconvenience.

What are the outputs of the following commands when this problem happens?

  $ collie cluster info -a 10.250.241.115
  $ collie cluster info -a 10.250.241.116
  $ collie cluster info -a 10.250.241.117
  $ collie cluster info -a 10.250.241.118
  $ collie cluster info -a 10.250.241.119


There are some fixes since 0.2.1.  Could you check whether the same problem
happens when you use the latest codes:

  git://sheepdog.git.sf.net/gitroot/sheepdog/sheepdog next


Thanks,

Kazutaka