On Sun, Sep 24, 2017 at 10:59 PM, Ben Bradley <listsbb(a)virtx.net> wrote:
On 23/09/17 00:27, Ben Bradley wrote:
> On 20/09/17 15:41, Simone Tiraboschi wrote:
>
>>
>> On Wed, Sep 20, 2017 at 12:30 AM, Ben Bradley <listsbb(a)virtx.net
>> <mailto:listsbb@virtx.net>> wrote:
>>
>> Hi All
>>
>> I've been running a single-host ovirt setup for several months,
>> having previously used a basic QEMU/KVM for a few years in lab
>> environments.
>>
>> I currently have the ovirt engine running at the bare-metal level,
>> with the box also acting as the single host. I am also running this
>> with local storage.
>>
>> I now have an extra host I can use and would like to migrate to a
>> hosted engine. The following documentation appears to be perfect and
>> pretty clear about the steps involved:
>>
https://www.ovirt.org/develop/developer-guide/engine/migrate
>> -to-hosted-engine/
>> <
https://www.ovirt.org/develop/developer-guide/engine/
>> migrate-to-hosted-engine/>
>> and
>>
https://www.ovirt.org/documentation/self-hosted/chap-
>> Migrating_from_Bare_Metal_to_an_EL-Based_Self-Hosted_Environment
>> <
https://www.ovirt.org/documentation/self-hosted/chap-
>> Migrating_from_Bare_Metal_to_an_EL-Based_Self-Hosted_Environment>
>>
>> However I'd like to try and get a bit more of an understanding of
>> the process that happens behind the scenes during the cut-over from
>> one engine to a new/hosted engine.
>>
>> As an experiment I attempted the following:
>> - created a new VM within my current environment (bare-metal engine)
>> - creating an engine-backup
>> - stopped the bare-metal engine
>> - restored the backup into the new VM
>> - ran engine-setup within the new VM
>> The new engine started up ok and I was able to connect and login to
>> the web UI. However my host was "unresponsive" and I was unable to
>> manage it in any way from the VM. I shut the VM down and started the
>> bare-metal ovirt-engine again on the host and everything worked as
>> before. I didn't try very hard to make it work however.
>>
>> The magic missing from the basic process I tried is the
>> synchronising and importing of the existing host, which is what the
>> hosted-engine utility does.
>>
>>
>> No magic up to now: the host are simply in the DB you restored.
>> If the VM has network connectivity and the same host-name of the old
>> machine you shouldn't see any issue.
>> If you changed the host-name moving to the VM, you should simply run
>> engine-rename after the restore.
>>
>
> Thank you for the reply.
> I tried this again this evening - again it failed.
>
> The host is present within the new engine but I am unable to manage it.
> Host is marked as down but Activate is greyed out. I can get get into the
> "Edit" screen for the host and on right-click I get the following options:
> - Maintenance
> - Confirm Host has been Rebooted
> - SSH Management: Restart and Stop both available
> The VMs are still running and accessible but are not listed as running
> under the web interface. This time however I did lose access to the
> ovirtmgmt bridge and the web interface, running VMs and host SSH session
> were unavailable until I rebooted.
> Luckily I left ovirt-engine service enabled to restart on boot so
> everything came back up.
>
> The engine URL is a CNAME so I just re-pointed to the hostname of the VM
> just before running engine-setup after the restore.
>
> This time though I have kept the new engine VM so I can power it up again
> and try and debug.
>
> I am going to try a few times over the weekend and I have setup serial
> console access so I can do a bit more debugging.
>
> What ovirt logs could I check on the host to see if the new engine VM is
> able to connect and sync to the host properly?
>
> Thanks, Ben
>
So I tried again to migrate my bare-metal host to a hosted VM but no luck.
The host remained in unresponsive state in the engine web UI and I was
unable to manage the host in anyway. Although all VMs continued to run.
I did capture some logs though.
From the new engine VM... engine.log
https://p.bsd-unix.net/view/raw/666839d1
From the host...
mom.log
https://p.bsd-unix.net/view/raw/ac9379f0
supervdsm.log
https://p.bsd-unix.net/view/raw/f9018dec
vdsm.log
https://p.bsd-unix.net/view/raw/bcdcdb13
Sorry but no one of that links is working right now.
The engine VM is complaining about being unable to connect to the
host,
though I can see from tcpdump communication is fine. I believe this is
backed up by the pings seen in mom.log
Though I can see the following in vdsm.log... [vds] recovery: waiting for
storage pool to go up (clientIF:569)
So I wonder if this is blocking the engine bringing the host up.
The host is running local storage, which I believe is a pretty recent
addition to ovirt. So I could see how trying to run an engine VM on a
host's local storage might cause weird issues.
Maybe I lost a piece here.
Are you trying to migrate an old 3.6 all-in-one setup (where ovirt-engine -
the manager - and vdsm - the agent on the host are running on the same
host) to hosted-engine in a single step?
In that case it will not work for sure since when you will power-down the
old engine machine you will also power-down your old host and so the new
engine running on the VM is not able to contact the host since the old host
is down.
In hosted-engine idea, the engine runs on a VM running on the host that its
managing.
In that case probably the most effective way is to:
within your old engine:
- create a datacenter with shared storage (please avoid using NFS in
loopback)
- add your new hosts there
- migrate all of your VMs to the new dataceter
- only now you can migrate to hosted-engine
I realise that there won't be HA with this setup, until I create my second
host and configure HA on the VM.
If I am unable to migrate from bare-metal -> engine VM then it doesn't
give me any confidence that I would be able to restore a setup from a
backup onto a bare-metal host and recover host state.
So is the only supported method of migrating from bare-metal engine to
hosted engine by...
1) Migrating to the appliance
2) Using a new host to migrate to VM
3) Using shared storage between hosts
Thanks, Ben
>
> The only detail is that hosted-engine-setup will try to add the host
>> where you are running it to the engine and so you have to manually remove
>> it just after the restore in order to avoid a failure there.
>>
>>
>> Can anyone describe that process in a bit more detail?
>> Is it possible to perform any part of that process manually?
>>
>> I'm planning to expand my lab and dev environments so for me it's
>> important to discover the following...
>> - That I'm able to reverse the process back to bare-metal engine if
>> I ever need/want to
>> - That I can setup a new VM or host with nothing more than an
>> engine-backup but still be able to regain control of exiting hosts
>> and VMs within the cluster
>>
>> My main concern after my basic attempt at a "restore/migration"
>> above is that I might not be able to re-import/sync an existing host
>> after I have restored engine from a backup.
>>
>> I have been able to export VMs to storage, remove them from ovirt,
>> re-install engine and restore, then import VMs from the export
>> domain. That all worked fine. But it involved shutting down all VMs
>> and removing their definitions from the environment.
>>
>> Are there any pre-requisites to being able to re-import an existing
>> running host (and VMs), such as placing ALL hosts into maintenance
>> mode and shutting down any VMs first?
>>
>> Any insight into host recovery/import/sync processes and steps will
>> be greatly appreciated.
>>
>> Best regards
>> Ben
>> _______________________________________________
>> Users mailing list
>> Users(a)ovirt.org <mailto:Users@ovirt.org>
>>
http://lists.ovirt.org/mailman/listinfo/users
>> <
http://lists.ovirt.org/mailman/listinfo/users>
>>
>>
>>
> _______________________________________________
> Users mailing list
> Users(a)ovirt.org
>
http://lists.ovirt.org/mailman/listinfo/users
>
_______________________________________________
Users mailing list
Users(a)ovirt.org
http://lists.ovirt.org/mailman/listinfo/users