Re: [Xen-API] Recovery of lost NFS VHD storage

[George Shuklin <george.shuklin@xxxxxxxxx>]

The simplest way is reboot all hosts in pool.

If you don't want to do this, there is very slow and annoying way to fix problem by hand without global reboot.

You need to kill all domains (xc.domain_shutdown()), and then kill every tapdisk process related to VHD files from NFS. This gonna be very hard, because of lagging NFS, but finally they all die. After that you can plug/unplug every pbd for damaged SR and reset locks by /opt/xensource/sm/ resetvdis.py

If this is too hard to do - reboot'em all.

PS Most brutal way to reboot host is execute command:

echo b >/proc/sysrq-trigger

no sync, no grace, no mercy, no shame, no delay. Just reboot.


On 15.04.2013 19:03, Michael Vistein wrote:

Hi,

Â

this morning I encountered a problem with XCP 1.6 in relation with a NFS storage repository.

Â

We are using a hardware pool with three identical servers, all are accessing a shared NFS VHD stroage repository on an external NFS server. This morning the NFS server crashed, therefore all VMs lost their hard drive and were more or less hanging.

Â

What is the official recovery method in this case? XenCenter still showed the SR as âconnectedâ, but a rescan of the SR failed. Directly on the console of the XCP I could not cd into the mountpoint due to âStale NFS handleâ. I wasnât able to unmount or remount the SR because of open files from the still running VMs. Shutting down or migrating VMs of course wasnât possible either.

The only solution I found was a hard reboot of all servers in the pool. Is there a better way for such a problem?

Â

Thanks in advance,

Â

Michael

_______________________________________________
Xen-api mailing list
Xen-api@xxxxxxxxxxxxx
http://lists.xen.org/cgi-bin/mailman/listinfo/xen-api

_______________________________________________
Xen-api mailing list
Xen-api@xxxxxxxxxxxxx
http://lists.xen.org/cgi-bin/mailman/listinfo/xen-api