[ovirt-users] problems migrating to hosted engine from bare metal

[cmc]

Hi Juan,

Thanks for the reply. I didn't have any reliable NFS storage, or any
Gluster or iscsi storage available to me, but I found some unused
space on the fibre channel storage that is used as the storage domain
for the cluster, which was very fortunate. So I created a LUN on that.
I also found out that storage for the hosted engine is not allowed to
be shared with anything else (not sure why) so I made a small 140GB
partition on that. I also found out that there are plans to be able to
migrate the Hosted Engine onto different storage later on (it is in
development?).

I was able to deploy without error, but now the hosted engine cannot
be migrated to other hosts due to an error, so I've sent an email
about that.

Kind regards,

Campbell

On Thu, Jun 22, 2017 at 4:24 PM, Juan Pablo <pablo.localhost at gmail.com> wrote:
> hosted engine lives on the hosted engine share you defined at the
> begining . it does not migrate later to another storage, if you want
> that you need to re-deploy.
> hows your hardware and logical setup? will it change ?
>
>
>
> 2017-06-13 10:59 GMT-03:00 cmc <iucounu at gmail.com>:
>> Hi,
>>
>> I created a new host to deploy a hosted engine, and then used a backup
>> from the bare metal engine and restored this, as per the procedure in:
>>
>> http://www.ovirt.org/documentation/self-hosted/chap-Migrating_from_Bare_Metal_to_an_EL-Based_Self-Hosted_Environment/
>>
>> Everything worked fine up until step 15 ('Continue setup') as the
>> script said the engine was not responding. I tried the reboot option
>> (option 3), but still it would not connect. So I could not do the
>> final step involving the internal CA, adding the host to an existing
>> cluster (of which there were two other hosts). I was able to connect
>> via vnc and ssh fine to the engine, and from here I could see that the
>> ovirt-engine service was up. I had to install the aaa-ldap extension
>> package to enable ldap auth separately however, but once done I was
>> able to log in, and it showed the old cluster as it was on the bare
>> metal engine. I added the host that I created the hosted engine on,
>> and it installed various packages and then I configured the network
>> and it looked fine, apart from the fact that I could not see a VM
>> named 'HostedEngine' in the list of VMs. I think however that this was
>> not a properly working setup, as the NFS storage I used to setup the
>> hosted engine became unavailable and I think this killed the hosted
>> engine, which caused it to reboot the host it was on. The hosted
>> engine has not come back since then, so I'm guessing it either isn't
>> properly set up for HA or it needs the NFS storage or something else
>> was not properly done by me in the setup. I've restarted the bare
>> metal engine for now as I needed it running for now.
>>
>>  My questions are:
>>
>> 1. My understanding is that the NFS storage is initially used to
>> create the hosted engine disk image, and is temporary, and that the
>> hosted engine later gets migrated to the storage used by the rest of
>> the cluster (which in my case is directly attached to the hosts via
>> fibre channel). I suspect that this did not happen. The bare metal
>> engine had some local ISO storage (on a hard disk local to it), which
>> will not be replicated to the hosted engine VM - will this cause a
>> problem for the deployment? I can create some new ISO storage later if
>> not.
>>
>> 2. What is the recommended way to recover from this situation? Should
>> I just run 'hosted-engine --deploy' again and try and find out what is
>> going wrong at step 15?
>>
>> I can probably get the disk image that was on NFS and mount it to find
>> out what went wrong on the initial deployment, or I can run the
>> deployment again and then get the log when it fails at step 15.
>>
>> Ovirt version was 4.1.2.2
>>
>> Thanks for any help,
>>
>> Cam
>> _______________________________________________
>> Users mailing list
>> Users at ovirt.org
>> http://lists.ovirt.org/mailman/listinfo/users