[sheepdog-users] Crash of a node

Valerio Pachera sirio81 at gmail.com
Tue Jun 18 09:28:48 CEST 2013


2013/6/18 Liu Yuan <namei.unix at gmail.com>:
> before you stopped the cluster, did you shutdown the guest?

Yes.

> From the log, some objects are missing, maybe you can first 'vdi check'
> it to see if really some objects are missing?

I remember I run the guest with 'cache=writebackup' when sheep had no -w option.
I noticed that, then I turned off the guest and restarted it with
cache=directsync fisrt, and without any option cache later.
I was getting I/O errors so I restarted the cluster using the cache
(-w for sheep and writeback for qemu).
Anyway, this may be confusing.
I have to make this repeatable.



More information about the sheepdog-users mailing list