[sheepdog-users] Problems with snapshot and cache

[Liu Yuan]

On Fri, Jul 12, 2013 at 12:38:29PM +0200, Valerio Pachera wrote:
> Hi all, I have a 3 node testing cluster.
> On this cluster I'm simulating the configuration I have in production.
> In production cache is not enabled.
> 
> I'm running a single guest where I've been writing lot's of data and
> now I was taking a snapshot (guest is running), but I get this error:
> 
> root at test004:~# collie vdi snapshot -s lug12 backup_data
> Jul 12 12:21:32 [main] do_read(297) failed to read from socket: -1,
> Resource temporarily unavailable
> Jul 12 12:21:32 [main] exec_req(405) failed to read a response
> Failed to write object 80c8d12800000000
> 

This means that the node is too much busy to reponse and the collie timeout.
was guest doing heavy IO while taking snapshot?

> root at test004:~# collie vdi list
>   Name        Id    Size    Used  Shared    Creation time   VDI id  Copies  Tag
>   backup       0   10 GB  2.4 GB  0.0 MB 2013-07-10 16:58   19093f
> 2
>   backup_data     0  1.0 TB  501 GB  0.0 MB 2013-07-10 16:55   c8d128
>    2
> 
> sheep.log
> Jul 11 08:36:19 [main] md_add_disk(146) /mnt/sheep/dsk02, nr 1
> Jul 11 08:36:19 [main] send_join_request(928) IPv4 ip:192.168.2.44 port:7000
> Jul 11 08:36:19 [main] for_each_object_in_stale(388) /mnt/sheep/dsk02/.stale
> Jul 11 08:36:19 [main] main(780) sheepdog daemon (version
> 0.6.0_94_ge70a103) started
> Jul 12 12:21:32 [gway 27528] push_cache_object(516) failed to push
> object Object is read-only
> Jul 12 12:21:32 [main] modify_event(151) event info for fd 23 not found

Hmm, there is a bug after restarting of sheep. Will fix it soon.

Thanks
Yuan