Re: [ovirt-users] Self-hosted engine: Host cannot activate - network issues

On Tue, Dec 29, 2015 at 4:15 AM, Alan Murrell <lists@murrell.ca> wrote:
I am attempting to install oVirt 3.6 on a CentOS7 host. it is a single server, so I am trying to install a self-hosted engine. I am able to install oVirt 3.5 self-hosted engine, but am having some problems doing it with 3.6.
Everything appears to go fine; the engine-setup completes successfully. On the host, I pressed"1" to indicate engine-setup was complete, and it can connect to the webadmin but the host never becomes operational.
I am able to log in to the webadmin, and indeed the host is in a non-operational state. When I click on the host then click on the "Virtual networks" tab, all my interfaces are showing a red arrow.
When I click on "Setup Host Networks", my "ovirtmgmt" network is unassigned. Believing this to be why my host is not operational (since it is under "Required"), I assigned it to my management interface, but the interfaces still remain red. In addition, the "ovirtmgmt" network does have a green arrow, but also the icon indicating "Out of sync". If I click on "Sync All Networks", I get the following error:
"Error while executing action SyncAllHostNetworks: Network is currently being used"
which makes sense, since "ovirtmgmt" is assigned/in use. I am unable to unassign ovirtmgmt.
Current status is this:
- All network interfaces are showing the red "down" arrow - "ovirtmgmt" is assigned to my management NIC, but indicating that it is out of sync - I am unable to unassign the "ovirtmgmt" network (or at least, if I do, the "OK" button is greyed out)
I could destroy the engine VM and go through the setup again, with the idea of once it is installed and waiting for the host to become operational, I could do a "Sync All Networks" and see if the networks turn green and go from there.
I wanted to see what sort of insight I could get from here first. I am not sure what logs would be useful to you, so let me know what you want to see and I and make them available (I will likely zip them up and post a link)
from host: /var/log/ovirt-hosted-engine-*/* (-setup and -ha) /var/log/vdsm/* from engine: /var/log/ovirt-engine/* /var/log/ovirt-engine/host-deploy/* If last one is empty, it means the engine did not manage to get host-deploy logs. Please check and try to copy what you find from /tmp of the host while waiting for host-deploy ("Waiting for the host to become operational").
Something else I could try if this becomes a "puzzler" is to do a 3.5.x self-hosted engine then perform an in-place upgrade to 3.6 and see if that works? I would rather try to help make a direct 3.6 install work, though...
Did you principally do exactly the same thing in 3.5 and 3.6? In both cases just had a simple network interface (no vlan/bonding/whatever) and input it when asked? What exact versions (OS and ovirt)? I also changed the subject. Best, -- Didi

On 28/12/2015 11:29 PM, Yedidyah Bar David wrote:
Did you principally do exactly the same thing in 3.5 and 3.6?
I did. I basically followed "Up and Running with oVirt 3.5", except I used 3.6 packages.
In both cases just had a simple network interface (no vlan/bonding/whatever) and input it when asked?
Well, I pre-created the 'ovirtmgmt' bridge, as I had to do that under 3.5 otherwise I ran into problems during setup. I can't recall offhand what the bug was (I posted about it a while back), but perhaps whatever it was has been fixed in 3.6, and I shouldn't pre-create the 'ovirtmgmt' bridge? After I sent my earlier e-mail on this issue, I played around with a couple things and actually got it up and working (the interfaces turned green, etc.). As a test, I rebooted the host, but now I can't even ping the host. it is at the office, so I won't have physical access to it until I go in in a few hours.
What exact versions (OS and ovirt)?
CentOS 7 minimal (after 'yum update', I think it it 7.2?). oVirt is 3.6.1, I believe (whatever version is installed form latest repo packages) Since I have lost access, I suspect something happened to the 'ovirtmgmt' interface. There is nothing on it that matters, so I may just do a fresh install, but without pre-creating the 'ovirtmgmt' bridge and see if the setup does it (which I understand itis supposed to?) Regards, Alan

On Tue, Dec 29, 2015 at 9:56 AM, Alan Murrell <lists@murrell.ca> wrote:
On 28/12/2015 11:29 PM, Yedidyah Bar David wrote:
Did you principally do exactly the same thing in 3.5 and 3.6?
I did. I basically followed "Up and Running with oVirt 3.5", except I used 3.6 packages.
In both cases just had a simple network interface (no vlan/bonding/whatever) and input it when asked?
Well, I pre-created the 'ovirtmgmt' bridge, as I had to do that under 3.5 otherwise I ran into problems during setup. I can't recall offhand what the bug was (I posted about it a while back), but perhaps whatever it was has been fixed in 3.6, and I shouldn't pre-create the 'ovirtmgmt' bridge?
It should have worked also in 3.5 in principle, not sure about your specific problems. In principle pre-creating the bridge should work, but is not commonly tested. Perhaps you also had issues with NetworkManager?
After I sent my earlier e-mail on this issue, I played around with a couple things and actually got it up and working (the interfaces turned green, etc.). As a test, I rebooted the host, but now I can't even ping the host. it is at the office, so I won't have physical access to it until I go in in a few hours.
What exact versions (OS and ovirt)?
CentOS 7 minimal (after 'yum update', I think it it 7.2?). oVirt is 3.6.1, I believe (whatever version is installed form latest repo packages)
Since I have lost access, I suspect something happened to the 'ovirtmgmt' interface. There is nothing on it that matters, so I may just do a fresh install, but without pre-creating the 'ovirtmgmt' bridge and see if the setup does it (which I understand itis supposed to?)
It is. Please open bugs if it does not work. Thanks and best regards, -- Didi

On 29/12/2015 12:17 AM, Yedidyah Bar David wrote:
In principle pre-creating the bridge should work, but is not commonly tested.
Perhaps you also had issues with NetworkManager?
After I do the minimal install I always stop and disable NetworkManager (and firewalld), anfd in the ifcfg-* files I set 'NM_CONTROLLED=no'. For good measure, I usually reboot also.
Please open bugs if it does not work.
Alright, I will do a fresh OS install on my server, not pre-create ovirtmgmt bridge, and run through the same steps as "Up and running with oVirt 3.5", except use 3.6 packages. If it gets stuck and does not create the ovirtmgmt bridge, I will file a bug. I file a bug with any other issues I encounter as well, if necessary. I will report back on on results. Regards, Alan

Everything seems to be good. The install created the "ovirtmgmt" bridge without any issue. At one point after I did the 'engine-setup', I thought the host got stuck, as it seemed to stay on "[INFO] Connecting to the engine" for quite a bit of time, but my "screen" session just got broken. I reconnected to the screen session, and was able to continue/finish the setup. Logging in to the webadmin indicates all seems to be fine (network interfaces are good, no errors, survives a reboot) Thanks for your insight on this! :-) Regards, Alan

On Tue, Dec 29, 2015 at 9:19 PM, Alan Murrell <lists@murrell.ca> wrote:
Everything seems to be good. The install created the "ovirtmgmt" bridge without any issue. At one point after I did the 'engine-setup', I thought the host got stuck, as it seemed to stay on "[INFO] Connecting to the engine" for quite a bit of time, but my "screen" session just got broken.
:-(
I reconnected to the screen session, and was able to continue/finish the setup.
For even less attended and hopefully quicker deploys, you can try ovirt-appliance. It's still not published on our repos, but you can use the last one from jenkins: http://jenkins.ovirt.org/job/ovirt-appliance_ovirt-3.6_build-artifacts-el7-x...
Logging in to the webadmin indicates all seems to be fine (network interfaces are good, no errors, survives a reboot)
Thanks for your insight on this! :-)
Thanks for the report! -- Didi
participants (2)
-
Alan Murrell
-
Yedidyah Bar David