cloudstack-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Carlos Reategui <car...@reategui.com>
Subject Re: Host failed issues
Date Wed, 03 Jun 2015 22:07:01 GMT
Continuing on this saga.  I did the following:

xe vm-list resident-on=bf9bb7a9-ee8e-46de-855d-9712ed037943
And I get several VMs that XenServer thinks are still running on the bad
host.  Once I get the host fixed and bring it back online what will happen
with all these?  Will XS try to start them?  CS thinks they are stopped.

On Wed, Jun 3, 2015 at 3:01 PM, Carlos Reategui <carlos@reategui.com> wrote:

>
>
> On Wed, Jun 3, 2015 at 1:52 PM, Andrija Panic <andrija.panic@gmail.com>
> wrote:
>
>> Hi Carlos,
>>
>> Im not familiar with Zen, but on KVM - you would "virsh destroy
>> i-xx-yy-VM"
>> and then edit the DB set VM state to Stopped.
>>
> Hmm... Interesting... "xe vm-list" does show the VM as running. However I
> was initially unable to stop it:
>
> # xe vm-shutdown name-label=i-3-201-VM force=true
>
> You attempted an operation which involves a host which could not be
> contacted.
>
> host: bf9bb7a9-ee8e-46de-855d-9712ed037943 (labxen03)
> I had to force the power state on the VM first:
>
> # xe vm-reset-powerstate vm=i-3-201-VM  force=true
>
> # xe vm-destroy uuid=dfb377f9-7f7b-bed7-88d9-a03b3a5d96e5
>
>
>> That should be sufficient to stop VM manually.
>>
> Yup.  Thank you
>
>
>>
>> Check from Zen Center if possible to shutdown VM...
>>
> XenCenter does not show the VM.  It only shows the ones on functional
> hosts.
>
>
>> If you still start VM for some reason on bad host - you can play with
>> last_host_id field in cloud.instance table (or similar name...), while the
>> VM is off, edit to point to some other host ID, and ACS should later try
>> to
>> start it on that  "new" host
>>
> This did it.  Table is vm_instance.  I have gone through the stopped VMs
> with the last_host_id of the bad host and changed it.  Hopefully this will
> do it.
>
> Thanks!
>
>
>>
>> Hope that helps
>>
>> On 3 June 2015 at 22:39, Carlos Reategui <creategui@gmail.com> wrote:
>>
>> > Hi All,
>> >
>> > I have a cluster with a pod of 6 XenServer 6.2 hosts on CloudStack
>> 4.4.0.
>> > One of the hosts stopped/rebooted suddenly and now can't find its disks
>> > (dell on the way).  Luckily the host was not the pool master but it did
>> > have the VR running on it so no instances could be created or started.
>> My
>> > network is a basic network without security groups.
>> >
>> > CS recognized that the host was down and also the VR, but it was unable
>> to
>> > move the VR to a different machine.  After destroying the VR from the
>> UI it
>> > came up on a different machine.  I would have expected CS to have moved
>> the
>> > VR automatically but maybe I don't have something setup correctly.
>> >
>> > In the hosts tab I put the failed host in maintenance mode.  I tried
>> > starting an instance that had previously been stopped.  CS started it on
>> > the bad host and reported that it was successful even though it was not.
>> > Why did it do this if the host was down and in maint mode??  Now I can't
>> > stop that instance or migrate it.  I took the host out of maintenance
>> mode
>> > and marked it as disabled instead.  I still can't stop the instance that
>> > thinks it is on the bad host.  Do I need to edit the DB to fix this
>> > instance?
>> >
>> > Any ideas?
>> >
>> > thanks,
>> > Carlos
>> >
>>
>>
>>
>> --
>>
>> Andrija Panić
>>
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message