[sheepdog-users] vdi problem

[Valerio Pachera]

First of all, everything got fixed!
I suspect troubles where related to a disk that got full on node 'sheepdog001'.
I didn't notice it before satturday (15/6/13).
I added a node (3T) and unplugged the full disk.
During the recovery, vdi check was failing, after it, vdi check was
fine on any of the vdi.
Also the host 'backup' doesn't show I/O error anymore!
Data was fine, I didn't need to fsck or anything else.

2013/6/17 MORITA Kazutaka <morita.kazutaka at gmail.com>:
> It would be great if
> you could look into the condition where it happens.

I'm keeping track of ram usage using script that run 'ps axv | grep sheep'.
Cluster recovery started about 12:00 of 15/06.
Notice the increase of used ram from 185592 (04:39:01)  390072 (04:40:01).
There's nothing on sheep.log after 'Jun 16 02:59:30'.

06/15/13 13:32:01 293130 102992
06/15/13 14:15:02 297218 105560
06/15/13 16:51:01 289022 107160
06/15/13 18:09:01 289022 107436
06/15/13 20:21:01 293126 107220
06/15/13 20:50:01 293126 107220
06/16/13 01:07:01 379158 119876
06/16/13 01:54:01 387354 120008
06/16/13 03:43:01 362766 119980
06/17/13 04:19:01 407858 164976
06/17/13 04:39:01 407858 185592
06/17/13 04:40:01 686394 390072
06/17/13 05:00:01 670002 386512
06/17/13 06:05:01 670002 386416
06/17/13 08:48:01 670002 386524

At 04:00 a backup is starting.
This backup uses rdiff-backup and it's output dir is on the guest 'backup'.
This backup wrote only 50M but it reads many small files.
May this have anything to do with ram used by sheep?

> For example, does
> the heavy memory consumption happen even when you disable object
> cache?

I'm going to stop the cluster for a few fix.
I'll try to rerun the backup described above with the use of cache an without.

Thank you.