[Xen-users] [XCP] VM Journal aborts with NFS SRs

["dwight at supercomputer.org" <dwight@xxxxxxxxxxxxxxxxx>]

 I have a question about how XCP deals with NFS failures.

Namely, when running an XCP Cloud using NFS Storage Repositories
for the VMs, is there a reliable mechanism to prevent an NFS
failure from causing filesystem journal aborts on the VMs?

For example, say one has a dozen VMs, scattered among a few
slaves, all of which use a single, simple NFS server for the SRs.
Then, on the NFS server, one either stops the NFS service, or
just simply pulls out the ethernet cable.

When this happens, Linux or BSD VMs will all eventually hit a
journalling error when they can no longer write to the Virtual
Disk. Windows may as well, but it appears to be rather silent,
which is a bit disconcerting.

This behavior is rather inconvenient, as the root filesystem (and
any others using the NFS SR) automatically get remounted read-only.
It requires either a reboot, or bit of work by hand, to clean up.
As well as potentially lost data.

Now, this is not unexpected behavior when using NFS. And indeed,
one sees what looks like this type of error pop up on the mailing
list from time to time.

However, the wonders of Virtualization allow us to pause a VM at
will.

So, I'm wondering if this ability, to detect the error and
pause all VM's associated with that SR, has been implemented
in XCP? And if so, how to enable it?

Thanks in advance,

        -dwight-

_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-users