install pacemaker on a proxmox node

[tino]

Hello

Is it possible to install pacemaker in a proxmox node?. I would like to use pacemaker for manaing resources like drbd . I am getting the following message while trying to install pacemaker on a proxmox node.


# aptitude install pacemaker

The following packages have unmet dependencies:
corosync-pve : Conflicts: corosync but 1.4.2-3 is to be installed.
The following actions will resolve these dependencies:

Remove the following packages:
1) clvm
2) corosync-pve
3) fence-agents-pve
4) libpve-storage-perl
5) openais-pve
6) proxmox-ve-2.6.32
7) pve-manager
8) qemu-server
9) redhat-cluster-pve
10) resource-agents-pve
11) vzctl


Please let me know
Thanks
Tino

 

[mir]

Why can't you use the pacemaker installed by proxmox (corosync-pve)?

 

[tino]

Mir

Thank you for the reply.

Corosync is a cluster communication system that should be present along with any cluster resource managers like pacemaker ? . Not sure what you meant by "pacemaker installed by proxmox".

 

[tom]

pacemaker is not used on Proxmox VE, we use rgmanager instead.

 

[tino]

Hi, Tom,

I have already installed a replicated DRBD storage on two of the nodes. My plan is to create a storage HA using DRBD and floating IP. So that the proxmox wont be aware of the HA in the storage. For that, i need to install pacemaker on the nodes. Is it possible to establish my aim using the rgmanager ? (DRBD storage replication using floating IP and pacemaker)

 

[mir]

A floating IP can be achieved this way by rgmanager:

<rm>
    <failoverdomains>
      <failoverdomain name="webfailover" ordered="0" restricted="1">
        <failoverdomainnode name="pve1"/>
        <failoverdomainnode name="pve2"/>
      </failoverdomain>
    </failoverdomains>
    <resources>
      <ip address="172.16.3.7" monitor_link="5"/>
    </resources>
    <service autostart="1" domain="webfailover" name="web-adm" recovery="relocate">
      <ip ref="172.16.3.7"/>
    </service>
</rm>

 

[tino]

Ok, But i need to manage not only the floating IP but also the DRBD service, NFS service. etc. Can i manage all that using rgmanager?

 

[dietmar]

Ok, But i need to manage not only the floating IP but also the DRBD service, NFS service. etc. Can i manage all that using rgmanager?

Proxmox VE primary purpose is to run VMs. If you want to setup a HA storage array using pacemaker, you are better off to use a separate cluster.

 

[mir]

DRBD: http://www.drbd.org/users-guide/ch-rhcs.html

NFS is a not matter. I have never been able to configure HA NFS and I think it is impossible. The reason is that NFS clients and servers establishes locks in kernel space which is prohibited to break from user space where rgmanager, and pacemake for that matter, runs. This means you will never be able to make a resource failover without the possibility of loosing data. Therefore NFS is unfitted for HA. If you need HA storage you will have to turn to GFS or Ceph.

 

[dswartz]

This is completely wrong.

 

[mir]

This is completely wrong.

Who or what are you referring to? Please be a little more verbal

 

[dswartz]

To the sweeping generalization about how HA and NFS don't and can't work. I'm sorry, but you obviously have no clue what you are talking about. There are quite a few vendors that provide enterprise HA/NFS solutions. Your explanation as to why you believe this (kernel locks and such) is nonsense. If you don't understand why, you clearly don't know how NFS is used in these scenarios, nor how failover works.

 

[mir]

If you are so knowledgeable it should take a fraction of your time to describe a working solution for HA NFS which guaranties that data is newer lost under resource failover?

 

[Raymond Burns]

I'm with @mirNot that it's not possible and can't happen, but saying someone is crazy in comparison, but don't give the comparison.When it comes to HA for NFS I only know of http://www.high-availability.com/zfs-ha-plugin/Other than that, I have not seen or heard of a working High Availability solution using ZFS. IF you are going to shed some light, then shed away. If not then...

 

[dswartz]

zfs was not in the discussion. he stated he didn't think ha nfs could work. maybe i was harsh, but i know a number of products which do precisely this. it is not my job to spend time and effort doing his/your homework. if you believe this is not possible, the onus is on you to explain why, and not using vague handwaving.

 

[wahmed]

zfs was not in the discussion. he stated he didn't think ha nfs could work. maybe i was harsh, but i know a number of products which do precisely this. it is not my job to spend time and effort doing his/your homework. if you believe this is not possible, the onus is on you to explain why, and not using vague handwaving.

I am sorry, but clearly "Helping Hands to Community" is misunderstood. It is not matter of doing homework for somebody or not, it is matter of being patient enough to explain something politely which could benefit the person who is asking along with the whole community.

No matter, I mainly use CEPH since it got HA by design. But i do know it used to be possible and may be still is, to setup NFS HA cluster of 2 nodes. (http://www.howtoforge.com/high_availability_nfs_drbd_heartbeat)

@Tino, are you trying to include Proxmox node in NFS HA cluster using Pacemaker? How about setting up separate NFS Cluster and just connect it with Proxmox with usual NFS share? Are you trying to save number nodes you need?

 

[liska_]

Hi Tino,
I was trying to make exactly this same configuration, but at the end I gave it up. It was so complicated to install pacemaker and I prefer more recommended and tested solution.
Anyway, maybe you could try gluster. It is no problem to install it and use it on pve node. There are howtos for installation on two servers and it performs better than nfs concerning reading speed. You can see a howto here http://www.jamescoyle.net/how-to/435-setup-glusterfs-with-a-replicated-volume-over-2-nodes and there are some other articles about performance on proxmox and about ceph as well.

 

[spirit]

NFS is a not matter. I have never been able to configure HA NFS and I think it is impossible.

It's working with netapp san for example.

(Don't known with linux, maybe it's need some nfs session transfert during ha)

 

[mir]

It's working with netapp san for example.

(Don't known with linux, maybe it's need some nfs session transfert during ha)

Yes, with netapp san and other similar solutions you can have HA but this is a NFS+ solution (NFS + added features). NFS in itself does not provide session transfer so a lost NFS resource will hang for a specified time (socket timeout + NFS timeout and retries). When resource is transferred you loose any current sessions and if the application layer does not identify this and is capable of rerunning the session you loose data. I have tried this with RDBMS' and web based applikation which all were affected by this.

 

[tino]

@symcomm - My plan was like to use DRBD replication with pacemaker and to attach that HA NFS on the proxmox. Does this work if i use separate storage server and install pacemaker inside them?

@liska - I have no idea about GLusterFS. Can i install glusterfs alongwith proxmox? Does it provide device level replication like DRBD? What are its pros when comapared to DRBD ?