[ovirt-users] Can't Add Host To New Hosted Engine - "Server is already part of another cluster"

[Yedidyah Bar David]

On Fri, Mar 23, 2018 at 4:10 AM, Adam Chesterton
<chesterton.adam at gmail.com> wrote:
> Hi Sahina and Yedidyah,
>
> Thanks for the information and offers of help. I am pleased to report that
> I've resolved the issue I had, thanks to the prompting your requests gave
> me, and everything is functional. I shall attempt to explain what happened
> and how I fixed it.

Glad to hear that :-) Thanks for the report!

You might want to open bugs about the misleading error messages/texts, or
anything else you think might have helped you more quickly understand
what your problem was and/or lead you in the right direction.

Best regards,

>
> When I looked at the Gluster peer status, the Host01 was rejected by Host02
> and Host03 (I did check this back at the start, but didn't check it again
> and things had changed). I followed the Gluster docs to fix the rejected
> peer
> (http://gluster-documentations.readthedocs.io/en/latest/Administrator%20Guide/Resolving%20Peer%20Rejected/).
>
> This then gave me a different error message when trying to add Host02 or
> Host03, "no available server in the cluster to probe the new server", which
> only further confirmed that it was a Gluster issue, as was suggested.
>
> After some hair-pulling and wondering, I finally discovered that, in Compute
>> Hosts > Host01 under the General tab, it was complaining that Gluster was
> not active (even though it was running). I clicked the action item link to
> resolve that, and oVirt appeared to start actually managing the Gluster
> service. I could then add my other hosts, import the existing storage
> domains, and everything appears good now.
>
> Thanks again for the assistance and prompting me towards the right places to
> help me resolve it.
>
> Regards,
> Adam
>
>
> On Thu, 22 Mar 2018 at 20:48 Sahina Bose <sabose at redhat.com> wrote:
>>
>> On Wed, Mar 21, 2018 at 12:33 PM, Yedidyah Bar David <didi at redhat.com>
>> wrote:
>>>
>>> On Wed, Mar 21, 2018 at 8:17 AM, Adam Chesterton
>>> <chesterton.adam at gmail.com> wrote:
>>> > Hi Everyone,
>>> >
>>> > I'm running a 3-host hyperconverged Gluster setup for testing (on some
>>> > old
>>> > desktops), and recently the hosted engine died on me, so I have
>>> > attempted to
>>> > just clean up my existing hosts, leaving Gluster configured, and
>>> > re-deploy a
>>> > fresh hosted engine setup on them.
>>> >
>>> > I have successfully got the first host setup and the hosted engine is
>>> > running on that host. However, when I try to add the other two hosts
>>> > via the
>>> > web GUI (as I can no longer add them via CLI) I get this error: "Error
>>> > while
>>> > executing action: Server XXXXX is already part of another cluster."
>>>
>>> This message might be a result of the host's participation in a gluster
>>> cluster,
>>> not hosted-engine cluster. Please share engine.log from the engine.
>>>
>>> Adding Sahina.
>>
>>
>> Yes, it does look like that.
>>
>> Can you share details of
>> # gluster peer status
>> from your 3 nodes
>>
>> And also the address of the first host in the oVirt engine and below from
>> the HE engine:
>>
>> # su - postgres -c "psql -d engine -c \"select * from gluster_server; \""
>>
>>>
>>> >
>>> > I've tried to find where this would still be configured on the two
>>> > other
>>> > hosts, but I cannot find anywhere.
>>>
>>> If it's only about hosted-engine, you can check /etc/ovirt-hosted-engine
>>> .
>>>
>>> You might try using ovirt-hosted-engine-cleanup, although it was not
>>> designed
>>> for such cases.
>>>
>>> >
>>> > Does anyone know how I can stop these two hosts from thinking they are
>>> > still
>>> > in a cluster? Or, does anyone have some information that might help, or
>>> > am I
>>> > going to just have to start a fresh CentOS install?
>>>
>>> If you do not need the data, a reinstall might be simplest.
>>> If you do, not sure what's your exact plan.
>>> You intend to rely on the replication? So that you reinstall one host,
>>> add it,
>>> wait until syncing finished, then reinstall the other? Might work, no
>>> idea.
>>>
>>> Best regards,
>>> --
>>> Didi



-- 
Didi