Create Failover-IP

[MasterTH]

Hi,

i',m currently on setting up a pve2.1 HA-cluster.
The way i wonna go is, using the storage that is build into the first two servers in the cluster and export it via nfs. The data is replicated through drbd. Two drbd-devices and the primary of each device is different, so one device has pve21-#1 as primary and the other pve21-#2. DRBD aso works quite fine. Compiled it from source and installed new userland version.

But know, i'm looking for an option to switch an failover-IP from pve21-#1 to pve21-#2 if pve21-#1 fails and set drbd-device to primary, mount this device and export the data. Heartbeat would be a good way, but i tried to install it and had a conflict with libcorosync..-pve. So i'm not really sure where to add the failover ip and the failover action.

If i get this thing working, i think this could be a cheap HA-Clustersetup without a need for additional two Storage-Servers.

Kind regards

 

[jerim]

I set up two storage servers using drbd. I used pacemaker/corosync with a failover IP. It works beautifully. Heartbeat seems to be on its way out but it has been in use for so long that there is way more documentation on it than corosync.

Being stand alone servers, the IP failover was no issue. In your setup, I understand that you want to run the box as a Proxmox node and storage. I am guessing that you would need to have two NICs in each box. One NIC to be used by Proxmox and the other NIC to be used for storage. Maybe you could use the same NIC with two different IP addresses, but that just depends on the number of users you have and how much usage you get. If you have any specific questions, be sure to let me know. I have a step by step guide that I used.

Also, check out the Linux Cluster Management Console (http://lcmc.sourceforge.net/). You can use it to setup DRBD and various services such as failover IP. However, I have found it best to use it as a visual chart of your setup and do all the actual work by command line.

 

[MasterTH]

Hi thanks for reply.

The real problem is that proxmox has a package installed that heartbeat uses. But the package is not named by the default way, because i think the pve-dev team added some specific features in it, its called lib...-pve (i don't know the name of that lib right now). So i can't install the heartbeat package

i'm currently dealing with rgmanager to manage the needed failoverip. An additional NIC is already installed, i'm not new to the topic "clustering" but i'm searching a way to implement an HA-Storage for VZ as well as for KVM machines on two proxmox servers.

Everything is running in my test-environment, but i got a problem with the nfs-services. i get a nfs-stale when the resource is switched to another host. After umount -l und remount the storage everything works fine, maybe thats the way i've to go.
Moving the nfs-directory (to avoid this stale) is not really good, because i'd like to implement Loadbalancing on Storage also. So i have on drbd thats master on nodea an the other drbd thats master on nodeb.


umm... loadbalancing in storage?..... i think i made a mistake in my setup. all storage load is replicated to the other node..... i'll try to move /var/lib/nfs to the drbd device so everything should work fine.


Maybe i'll give a look into my configuration

 

[dietmar]

The real problem is that proxmox has a package installed that heartbeat uses.

There is no need to use heartbeat.