[PVE-User] I try hard but...

[Emmanuel Kasper]

Hi

On 10/26/2015 02:25 PM, Dmitry Petuhov wrote:
> There's issue with NFS: if you try to send over it more than network can deal 
> with (100-120 MBps for 1-gigabit), it imposes several-second pauses, which are 
> being interpreted like hardware errors. These bursts may be just few seconds 
> long to trigger issue.
> 
> You can try to limit bandwidth in virtual HDD config to something like 60-70 
> MBps. This should be enough for 1-gigabit network.
> 
> But my opinion is that it's better to switch to iSCSI.


Is that something that you can easily reproduce  ? ( dd with conv=fsync
or other tool)

I do have heard about transient problems from NFS Servers at some
hosters, when running vzdump on a NFS target, the backup would hang at
some point because the NFS connection was lost.

Up to know I though this was caused by some kind of network throttling
done by the hoster, which was causing the traffic to be blocked longer
that the NFS timeout.