Need help about PVE Cluster

[samontetro]

For several weeks I'm trying to build a PVE cluster (not H.A. at this time) with 2 nodes for testing (so I'm allowed to break everything!). I've understood that 3 nodes is better but I've only 2 and I read that it is possible to do this.
My documentation is the wiki and "Proxmox High Availability" book by Simon M.C. Cheng.

My problem today is that if I install a lxc based VM on host A it does not exist on node B and I do not understand how to make it available to manualy stop the VM on node A and start it on node B (no automatic migration at this test level).

Node A and B are same hardware, running proxmox 4.2
Cluster is created and I can manage the two nodes from the same web interface
I've setup a volume group on a drbd8 config primary/primary (working) where I put the root file system of my VM and I can see its LV on the 2 nodes with the same path (lvdisplay).

But VM is created on node A and doesn't exist on node B. I understand I should not install it twice.
I've read about pmxcfs replicating configs between nodes and I think I've missed something there...

Thanks for your advices or links to some related documentation I should have read.

Patrick

 

[udo]

For several weeks I'm trying to build a PVE cluster (not H.A. at this time) with 2 nodes for testing (so I'm allowed to break everything!). I've understood that 3 nodes is better but I've only 2 and I read that it is possible to do this.
My documentation is the wiki and "Proxmox High Availability" book by Simon M.C. Cheng.

My problem today is that if I install a lxc based VM on host A it does not exist on node B and I do not understand how to make it available to manualy stop the VM on node A and start it on node B (no automatic migration at this test level).

Hi Patrick,
if the CT (or VM) has all data on shared (or replicated storage) you need only to move the config to see the CT/VM on the other node.
Look in /etc/pve in nodes ans also see the links qemu-server and lxc. The links are different on both nodes.

Here is the content of /etc/pve of an single node:

/etc/pve
├── authkey.pub
├── datacenter.cfg
├── local -> nodes/pve-a
├── lxc -> nodes/pve-a/lxc
├── nodes
│  └── pve-a
│  ├── lrm_status
│  ├── lxc
│  ├── openvz
│  ├── priv
│  ├── pveproxy-ssl.key
│  ├── pveproxy-ssl.pem
│  ├── pve-ssl.key
│  ├── pve-ssl.pem
│  └── qemu-server
│  ├── 100.conf
│  ├── 101.conf
│  ├── 102.conf
│  ├── 103.conf
│  ├── 104.conf
│  ├── 105.conf
│  ├── 110.conf
│  └── 111.conf
├── openvz -> nodes/pve-a/openvz
├── priv
│  ├── authkey.key
│  ├── authorized_keys
│  ├── known_hosts
│  ├── lock
│  ├── pve-root-ca.key
│  ├── pve-root-ca.key.org
│  └── pve-root-ca.srl
├── pve-root-ca.pem
├── pve-root-ca.pem.org
├── pve-www.key
├── qemu-server -> nodes/pve-a/qemu-server
├── storage.cfg
├── user.cfg
└── vzdump.cron

an move could be "mv /etc/pve/nodes/pve-a/qemu-server/100.conf /etc/pve/nodes/pve-b/qemu-server/" (if I had an second node which named pve-b).

BUT... if one node die, you don't have quorum - mean that /etc/pve is read only - so an move isn't possible. This is the reason, why three nodes are recommended.
If you be sure, that the other node is dead, you can reach quorum with "pvecm expected 1".

Node A and B are same hardware, running proxmox 4.2
Cluster is created and I can manage the two nodes from the same web interface
I've setup a volume group on a drbd8 config primary/primary (working) where I put the root file system of my VM and I can see its LV on the 2 nodes with the same path (lvdisplay).

You mean drbd9?!

Udo

 

[samontetro]

Thanks Udo for this explanation. I was thinking that installing a new VM will lead to show it on all the nodes, even if it will run only on one node.
With "pvecm expected 1" I can stop the node runing my VM and restart it later on the other node by hand. It works. Now i've to investigate split brain events with such a config.

I've downgraded drdb back to de 8 version as many people seams to have trouble with the latest version and I do not need the new functionalities of drbd9. I was not able to setup properly my config with drbd9 and it was "out of the box" with drbd8. I'm familar with older drbd versions.

I'm trying to negociate a proxmox training session as version 4 introduces many new concepts.