8:38 p.m.
Hi all,
I used the oVirt installer via cockpit to setup a hyperconverged cluster on 3 physical
hosts running Red Hat 8.3. I used the following two resources to guide my work:
-
https://www.ovirt.org/documentation/gluster-hyperconverged/chap-Deploying...
- https://blogs.ovirt.org/2018/02/up-and-running-with-ovirt-4-2-and-gluster...
As I mentioned in a previous email a week ago, it appears as if the wizard successfully
setup gluster on all 3 of the servers, and the wizard also successfully setup the ovirt
engine on the first host.
However, the oVirt engine only recognizes the first host, and only the first host's
physical resources are available to the cluster.
So I have Gluster 8 installed on the 3 hosts (RHEL 8), and oVirt 4.4.5 installed on the
first host, along with the ovirt engine VM.
I run `gluster peer status` from the first node, can confirm that the other two physical
hosts are healthy:
(example.com is replacing actual domain below)
[root@cha1-storage ~]# gluster peer statusNumber of Peers: 2Hostname:
cha2-storage.mgt.example.comUuid: 240a7ab1-ab52-4e5b-98ed-d978f848835eState: Peer in
Cluster (Connected)Hostname: cha3-storage.mgt.example.comUuid:
0563c3e8-237d-4409-a09a-ec51719b0da6State: Peer in Cluster (Connected)
I am now trying to get the other two hosts added to the Engine. I navigate to Compute
-> Hosts, and click New, fill in the details (hostname, root password, etc..) and begin
the installation on the additional hosts.
It keeps failing.
Checking the error logs in /var/log/ovirt-engine/host-deploy on the Engine VM, I see the
following near the bottom:
"msg" : "Failed to download metadata for repo 'ovirt-4.4-epel':
Cannot prepare internal mirrorlist: Curl error (7): Couldn't connect to server for
https://mirrors.fedoraproject.org/metalink?repo=epel-8&arch=x86_64&am...
[Failed to connect to mirrors.fedoraproject.org port 443: No route to host]",
I can confirm that the Engine and each of the hosts are able to get to
mirrors.fedoraproject.org:
I've run the following on both the Engine, as well as each of the hosts (the first
host where everything is installed, as well as the 2nd host where I'm trying to get it
installed):
[root@ovirt-engine1 host-deploy]# curl -i https://mirrors.fedoraproject.org
HTTP/2 302
Note, that this traffic is going out the management network.
That may be an important distinction -- keep reading.
So this does lead me to another issue that may or may not be related.
I've discovered that, from the RHEL 8 host's perspective, the public facing
network is unable to get out to the internet.
Management (Gluster & the Engine VM) is on: 10.1.0.0/24
Each of the hosts are able to ping each other and communicate with each other.
Each of the hosts are able to ping 8.8.8.8 whenever the frontend network interface is
disabled (see below).
The frontend network is on: 192.168.0.0/24
Each of the hosts are able to ping each other on this network
But this network isn't able to get out to the internet (yet).
Obviously I need to fix the routing and figure out why the 192.168.0.0/24 network is
unable to reach the internet.
But shouldn't all the traffic to install the ovirt functionality onto the 2nd host go
out the management network?
So to summarize, I have a few questions:
- Why didn't the wizard properly install the ovirt functionality onto all 3 hosts to
begin with when I did the initial installation?
- From the physical host's perspective, what should be the default route? The
internal management, or the front-end network?
- Is the following statement accurate?
- The ovirt-engine's default route should be management -- it doesn't even have
a front-end IP address.
- Why would the ovirt engine fail to install, and indicate that it cannot get to
mirrors.fedoraproject.org, when clearly it can?
- Again, the ovirt-engine VM is able to curl that URL and gets a valid http response.
- Any other tips or suggestions on how to troubleshoot this?
Sent with ProtonMail Secure Email.