Bad news.
Bringing everything back online again after cold booting, it appears that the old & complicated vlan settings came back, yet again, on two of the hosts.

Every time I boot the infrastructure, these old settings come back.

I think at this point, I'm going to just fresh install everything, because I can't keep running into this issue, and I'm starting to run really low on available time.

My plan is to have the hardware deployed into the datacenter on Friday. 
I don't have a whole lot of work put into the VMs themselves (thankfully).


Sent with ProtonMail Secure Email.

‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐
On Monday, April 12, 2021 5:25 PM, David White via Users <users@ovirt.org> wrote:

Thanks to you and Did, both.
Very helpful.

I've whittled everything down to their most basic settings on the hosts, and did what you suggested, to try to stabilize everything on the most basic network settings.
That did the trick, and everything is now able to get to everything now.

I've also decided to simplify my datacenter installation, and go with a traditional NAT-based architecture.

I have two routers, so I'll stick half the IPs on 1, and half the IPs on the other, with VPN capability on each, so that I and my team can quickly fail customers over from 1 router to the other, in case of an uplink or a router issue.

Simplicity is king, and I think I was trying to make this way too complicated.

Everything is 100% now, and I'll work on the NAT-based architecture over the next few days.

Thanks again,
David



Sent with ProtonMail Secure Email.

‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐
On Monday, April 12, 2021 1:49 PM, Vincent Royer <vincent@epicenergy.ca> wrote:

In the beginning, I made this same mistake.  I setup all my networks, bonds, vlans, and tried to deploy the engine from cockpit.  This never worked properly.  Eventually I decided to leave the networks as simple as possible - just a static IP on a single interface for management, and same for Gluster. Then I setup the networks the way I wanted inside the 'setup host networks' window, and then 'sync all networks'.  This worked much better, and now it's pretty easy to add VLANs as required. 

image.png







On Sun, Apr 11, 2021 at 4:47 AM David White via Users <users@ovirt.org> wrote:
Thanks for the response here.
Unfortunately, things are still not 100% stable after I performed that host upgrade.
It appears to me that 1 of the hosts keeps booting back with the old management vlan (VLAN 1) instead of the new vlan (VLAN 10).

I'm able to (mostly) fiddle around with it and get it back online, but it seems like every reboot, vlan 1 comes back, and breaks connectivity with the other two hosts.
I also didn't realize, until late yesterday afternoon, that you could use the oVirt Manager web UI to configure each host's network settings.

This whole time, I've been trying to use nmcli, nmtui, and manually editing /etc/sysconfig/network-scripts/ files by hand.
When I found the network settings inside of the Engine / manager web UI, I discovered that, for example, oVirt Engine manager saw that the new vlan (VLAN 10) was "unmanaged" by the engine.
Question: Would you recommend that all network settings be modified by the oVirt engine instead of the manual process on the OS?

Question: Is it possible to setup a vlan inside the engine without an IP address being assigned to each of the physical hosts?
I was really hoping to setup VLAN 100 with public IP addresses, and use layer 2 switching to send that traffic into the oVirt cluster.

Here is a screenshot overview of what I want my environment to look like, logically. You'll note that I was going to put VLAN 20 and VLAN 100 onto the same host physical interface.
This is what I - and I think the oVirt documentation - refers to as the front-end traffic.
VLAN 10 is/was going to be on its own interface going to the 10Gbps switch.

Question: Do you see anything "wrong" with this picture? Are there ways I can / should change it to improve?

As for the /etc/hosts files, I'm actually doing that.
However, as I'm typing this, I realized that I never defined the Engine IP address in the hosts, nor do I put anything inside the Engine's /etc/hosts file.
Question: Perhaps this was part of my problem, when DNS connectivity was not working. Thoughts?

Thanks again,
David

Screenshot from 2021-04-10 17-21-39.png


Sent with ProtonMail Secure Email.

‐‐‐‐‐‐‐ Original Message ‐‐‐‐‐‐‐
On Sunday, April 11, 2021 3:05 AM, Yedidyah Bar David <didi@redhat.com> wrote:

On Sat, Apr 10, 2021 at 1:14 PM David White via Users <users@ovirt.org> wrote:
This is resolved, and my environment is 100% stable now.

Glad to hear that, thanks for the report!
 

Or was, until I then used the engine to "upgrade" one of the hosts, at which point I started having problems again after the reboot, because the old vlan came back.
I'll finish getting things stabilized today, and hopefully won't run into this again.

I've been turning things on and off quite a bit, because they aren't in a proper data center (yet) and are just sitting here in my home office.
So I'm sure shutting them down and turning them back on fairly often hasn't helped the situation.

I initially had a few issues going on:
  1. I of course first broke things when I tried to change the management vlan
  2. Aside from my notes below and the troubleshooting steps I went through, yesterday, I had forgotten that connectivity to the DNS server hadn't been restored. Once I got DNS operational, the engine was able to see two of the hosts, and finally started showing some green.
  3. I then went in and ran `hosted-engine --vm-stop` to shutdown the engine, and then I started it again... and viola. The last remaining problematic host came online, and a few minutes later, the disks, volumes, and datacenter came online.
  4. I think part of my problem has been this switch. I purchased a Netgear GS324T for my frontend traffic. But I've also needed to put my backend traffic onto some temporary ports on that switch until I can get a VM controller setup that will run my other switch, a Ubiquiti US-XG-16 for my permanent backend traffic. The Netgear hasn't been nearly as simple to configure as I had hoped. The vlan behavior has also been inconsistent - sometimes I have vlan settings in place, and things work. Sometimes they don't work. It has also been re-assigning a of the vlans occasionally after reboots, which has been frustrating. I'm close to being completely done configuring the infrastructure, but I'm also getting increasingly tempted to go find a different switch. 
Lessons learned:
  1. Always make sure DNS is functional
    1. I was really hoping that I could run DNS as a VM (or multiple VMs) inside the cluster.
    2. That said, if the cluster and the engine won't even start correctly without, then I may need to run DNS externally. I'm open to feedback on this.
      1. I have 1 extra U of space at the datacenter reserved, and I do have a 4th spare server that I haven't decided what to do with yet. It has way more CPU and RAM than would be necessary to run an internal DNS server... but perhaps I have no choice. Thoughts?

You can also have the IP addresses of the engine and hosts in /etc/hosts of all machines (engine and hosts) - then things should work fine. It does mean you'll have to manually maintain these hosts files somehow.
 

  1. Make sure your vlan settings are correct before you start deploying the hosted engine and configure oVirt.

Definitely. As well as making sure that IP addresses (and netmasks, routes, etc.) are as intended and working, name resolution is correct (DNS or /etc/hosts), etc. .
 

  2. If possible, don't turn off and turn on your servers constantly. :) I realize this is a given. I just don't have much choice in the matter right now, due to lack of datacenter in my home office. 

While definitely not recommended, in principle this should be harmless. If you find concrete reproducible bugs around this, please report them (with clear accurate details - just "I turn off and on my hosts and things stop working" is not helpful, obviously...).

Thanks again and best regards,




Users mailing list -- users@ovirt.org

_______________________________________________
Users mailing list -- users@ovirt.org
To unsubscribe send an email to users-leave@ovirt.org
Privacy Statement: https://www.ovirt.org/privacy-policy.html
oVirt Code of Conduct: https://www.ovirt.org/community/about/community-guidelines/
List Archives: https://lists.ovirt.org/archives/list/users@ovirt.org/message/NQQPZMLIDCVYH3GC4VGXELRLS7RSYDJ7/