Re: [Xen-API] Host Down VDI Locked
This is good if the environment is a test one and no one bother
shutting down all VMs just to regain lost VM(s).
Another important information is missing in your steps is that one
should have a "Virtual Machine Metadata" already backed up else it will
loose his SR permanently.
XCP 1 rc 3 has solved this issue and Vms can safely boot up after a
On Thursday 24 February 2011 04:28 PM, Chris Percol wrote:
Here are some notes I made on recovering from vdi not
available after a power failure on our iscsi san. Just for reference
for others as you seems to be sorted anyway.
###Fix VDI Problems after Power Failure###
#Steps that fixed 'vdi not available', sr backend failure
xe pbd-list sr-uuid=<uuid>
xe pbd-unplug uuid=<uuid>
Storage > New Storage
Restore Virtual Machine Metadata
On Wed, Feb 23, 2011 at 9:09 PM, Rushikesh <rushikesh.jadhav@xxxxxxxxxx>
I am using and studying XCP from past 2 month and have got my self a
little acquainted with its architectural working.
My setup: 3 XCP 1beta hosts and shared iSCSI storage. ( not the rc3 )
I have just started getting my hands dirty on code. (Python is very new
I had a case when my host was hard rebooted while VM on it were running.
And now all are refusing to start with Exception "VDI is unavailable".
So, when I started reading the traceback more closely I found that the
exception is generating at
File "/opt/xensource/sm/blktap2.py", line 1230, in _add_tag
if sm_config.get("writable") == "true":
raise util.SMException("VDI %s already attached RW" % vdi_uuid)
I compared a crashed VDI and non crashed and saw that the crashed VDI
still has a writable : true in sm-config.
# xe vdi-param-list uuid=03dfd013-bdd0-4c76-b099-4ad30d4018b6 (
Crashed VDI )
sm-config (MRO): writable: true; vhd-parent:
# xe vdi-param-list uuid=f701575e-b399-4c02-a930-e3d0dee87b6e (
Normal VDI )
sm-config (MRO): vhd-parent: 545b2f45-552a-474c-8b4f-f6079907fd21;
Then I started reading the code but Im getting little lost and I dont
want to make xe database corrupt. Recreating SR is a solution with a
risk of getting all VM rebooted.
So, I tried same scene with XCP1 and all VM were able to get
successfully boot. Nice work guys :)
I really want to understand how you did it and was I on correct path...
I really want to get into some development of XCP but my first problem
is already solved ;)
Thanks for patience and expecting a positive reply.
xen-api mailing list
xen-api mailing list