[ovirt-users] Migrating (or Deploying) Hosted Engine to second host from AMD -> Intel

[Roy Golan]

On Mon, Dec 21, 2015 at 9:50 AM, Josh <josh at bnhtech.com> wrote:

> Hi,
>
> I am currently running a Ovirt 3.6 host + Hosted Engine (both Centos 7) on
> an AMD based computer (AMD_HOST). I recently purchased a second server
> which is running an Intel processor and want to move all the VM's + Hosted
> Engine to this second server (INTEL_HOST). I was wondering what the proper
> procedure to do this would be. I have attempted a few things but nothing is
> successfully working. I have added and activated the INTEL_HOST through the
> Hosted Engine UI but since the AMD_HOST is AMD based and the second is
> Intel based I had to create a separate cluster (INTEL_CLUSTER). I can move
> all VM's to this new cluster but have not figured out how to get a Hosted
> Engine working on INTEL_HOST.
>
> Attempting to force migration of the Hosted Engine to INTEL_CLUSTER fails
> with the following:
> Cannot migrate VM. There is no host that satisfies current scheduling
> constraints. See below for details:
> The host INTEL_HOST did not satisfy internal filter HA because it is not a
> Hosted Engine host..
>
> Attempting to deploy a second hosted engine also runs into errors
> (hosted-engine --deploy). After downloading the answers.conf file from
> AMD_HOST, setup fails because of a different process type/family. I
> manually downloaded the answers.conf file and changed the processor type
> but setup validation fails because the path to the hosted engine metadata
> does not exist.
>
> hosted-engine --deploy --config-append=answers.conf
> ...
>  [ ERROR ] Failed to execute stage 'Setup validation': failed to read
> metadata: [Errno 2] No such file or directory:
> '/rhev/data-center/mnt/<nfs_mount>/<he_uuid>/ha_agent/hosted-engine.metadata'
> [ INFO  ] Stage: Clean up
>
> This metadata file points to /var/run/vdsm/storage/<he_uuid>/<more
> uuids...> but the he_uuid folder does not exist. I suspect this path
> doesn't exist because the storage domain is mapped for AMD_CLUSTER's hosted
> engine and cannot be referenced by another cluster. vdsClient does list the
> storage domain info though on the INTEL_HOST.
>

There shouldn't be any difference in paths between amd and intel. Please
share the vdsm log from that intel host during the setup failure

>
> vdsClient -s 0 getStorageDomainInfo <he_uuid>
>         uuid = <he_uuid>
>         version = 3
>         role = Regular
>         remotePath = <nfs_path>
>         type = NFS
>         class = Data
>         pool = []
>         name = hosted_storage
>
> The next steps I was going to take was to backup the hosted engine,
> setting up INTEL_HOST, and restoring the backed up hosted engine but I'm
> not sure if this will work since it will be restored on a different host.
> The next option was to just export all VM's, setup INTEL_HOST and a new
> hosted engine from scratch, and import all VM's, though I would prefer not
> to have to do this.
>
> Does anyone have any recommendations on how to make this move? I would
> simply like to move my Ovirt installation onto the INTEL_HOST. I thought I
> would be able to add the second host, migrate all data over, and remove the
> first host, but this procedure is getting hung up on dealing with the
> hosted engine.
>
> Thanks,
> Josh
>
> _______________________________________________
> Users mailing list
> Users at ovirt.org
> http://lists.ovirt.org/mailman/listinfo/users
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/users/attachments/20151222/16e0b3a9/attachment-0001.html>