Created cluster - Node IP is 127.0.0.1

[h9k]

Hi there!

I have two machines which are connected with a private VLAN.
I properly set up the IP in the /etc/hosts file to be 172.16.0.2, but after I created the cluster with the

pvecm create clustername

and then did

pvecm status

it says

Node addresses: 127.0.0.1

which I guess is wrong. When I try to add the other machine to the cluster I type

pvecm add 172.16.0.2

and after first accepting the ssh key (added to known hosts) I get this error:

unable to copy ssh ID

I suppose this has to do with the IP of the main node being wrong.
The question now is: how do I fix that?

 

[udo]

Hi,
how looks your /etc/hosts?

I assume it's not an standard-installation? Because than should it work.
See alos in the wiki: http://pve.proxmox.com/wiki/Install_Proxmox_VE_on_Debian_Wheezy

An correct entry looks like this:

127.0.0.1 localhost.localdomain localhost
192.168.6.177 proxmox-6-177.proxmox.com proxmox-6-177 pvelocalhost

In this case "hostname -a" shows pvelocalhost.

Udo

 

[h9k]

Hi Udo,

here is my /etc/hosts
127.0.0.1 localhost.localdomain localhost
172.16.0.1 nebula.fulldomain.tld nsXXXXXX.ip-XX-XX-XX.eu nebula nsXXXXXX
172.16.0.2 clavius.fulldomain.tld nsYYYYYY.ovh.net clavius nsYYYYYY

Hostname -a shows "nsXXXXXX.ip-XX-XX-XX.eu nebula nsXXXXXX" for nebula and "nsYYYYYY.ovh.net clavius nsYYYYYY" for clavius.

Both hosts have the same hosts file. The strange thing is that when I created the cluster on nebula, the IP was detected fine. But as nebula is the new server without any containers yet and clavius is production, I reset the cluster on nebula and created in on clavius. But for some reason there it got the 127.0.0.1.

I have two OVH servers which were installed with the Proxmox image from OVH. Clavius was installed with v2 and at some point upgraded to v3, Nebula is a fresh v3 install.
They are connected via VLAN over the secondary network devices and the hosts ping each other fine, also SSH connections do work.

 

[udo]

Hostname -a shows "nsXXXXXX.ip-XX-XX-XX.eu nebula nsXXXXXX" for nebula and "nsYYYYYY.ovh.net clavius nsYYYYYY" for clavius.

Hi,
add pvelocalhost as alias in /etc/hosts.

Udo

 

[h9k]

Ok I did so and restarted pve-cluster, but pvecm status still says 127.0.0.1 for node IP.

 

[m.ardito]

Ok I did so and restarted pve-cluster, but pvecm status still says 127.0.0.1 for node IP.

Imho I guess it's best , if you can, to try the much simpler path of reinstalling nodes using iso installer, and provide it final ip addresses, and fqdn where asked there...

in a complex system as pve cluster there's more than /etc/hosts file to be setup.

Marco

 

[h9k]

Well I have production containers running on one box, I cannot just reinstall it. If the cluster was working I could migrate the containers, so I'm stuck here.
It sure must be possible to change the node IP somehow or reset something without affecting uptime of the running containers.

 

[m.ardito]

Well I have production containers running on one box, I cannot just reinstall it. If the cluster was working I could migrate the containers, so I'm stuck here.

Imho, if you really care about your production machines, don't try to play with networks setting on a cluster system you don't really know: ask before, and test before. Anyway, now you could backup/restore on working nodes, probably after having setup temporary alternative services on working nodes. (and also ask for pve paid support, if your business can't allow such alternative ways...)

Marco

 

[udo]

Ok I did so and restarted pve-cluster, but pvecm status still says 127.0.0.1 for node IP.

Hi,
I assume it's goes wrong during the creation of the cluster with the missing pvelocalhost entry.
See in /etc/pve/cluster.conf which name are used.

The official way is an new installation, but there are an thread in the forum, which files are must removed to "recluster" an node.

Udo

 

[h9k]

Yeah I just "reclustered" the clavius node after adding pvelocalhost to /etc/hosts.
Weirdly though it still assigned 127.0.0.1 as node IP.
In /etc/pve/cluster.conf the node name is "clavius". If I ping clavius I get the correct 172.16.0.2 IP.

 

[udo]

Yeah I just "reclustered" the clavius node after adding pvelocalhost to /etc/hosts.
Weirdly though it still assigned 127.0.0.1 as node IP.
In /etc/pve/cluster.conf the node name is "clavius". If I ping clavius I get the correct 172.16.0.2 IP.

Hi,
you have also restarted the pve-services?

Udo

 

[h9k]

Yes, I restarted all I could find: pvebanner pve-cluster pvedaemon pve-manager pvenetcommit pveproxy pvestatd

 

[h9k]

No matter what I try, it just doesn't work.
Would this be something you could fix if I had a basic subscription plan?

 

[m.ardito]

No matter what I try, it just doesn't work.
Would this be something you could fix if I had a basic subscription plan?

I cannot speak for pve support, but basic subscription includes some support tickets, afaik, see http://www.proxmox.com/proxmox-ve/pricing

Marco

 

[kickinz1]

Hi,

I had the same problem, I have edited /etc/hosts, /etc/pve/corosync.conf to the correct /24 address, then I have restarted corosync and pve-cluster services. I encoutered also a problem with authorized kzys already present when wanting to re-add nodes. I used pvecm add IP -force, and my cluster is now up and running.

Just in case someone encountered the same pb.