[Gluster-users] Gluster 3.4 on RHCS with OCF resource agents

[Emir Imamagic]

Hello,

by looking at the source code I managed to find answer to my second 
question - yes :)

In order to run multiple glusterd on the same host one needs to provide 
option transport.socket.bind-address in the volfile, e.g. 
/etc/glusterfs/glusterd-1.vol:
---
volume management
     type mgmt/glusterd
     # don't forget to change the working-directory
     option working-directory /var/lib/glusterd-1
     option transport-type socket,rdma
     option transport.socket.keepalive-time 10
     option transport.socket.keepalive-interval 2
     option transport.socket.read-fail-log off
     option transport.socket.bind-address XX.YY.ZZ.AA
end-volume
---
In addition to different address and working directory one has to define 
different PID file when starting the service:
  glusterd -f /etc/glusterfs/glusterd-1.vol -p /var/run/glusterd-1.pid

Finally when using the CLI one has to specify to which glusterd to connect:
  gluster --remote-host=XX.YY.ZZ.AA
After that business as usual by using virtual IPs.

With this knowledge I think I'll be able to make the OCF resource agents 
work. Even if not I can always use service resource agent with two 
modified init.d scripts.

Cheers,
emir

On 30.10.2013. 12:58, Emir Imamagic wrote:
> Hello,
>
> Gluster 3.4 provides OCF resource agents which should enable nice
> integration with RHCS. We would like to deploy gluster on two RHCS nodes
> with storage provided via iSCSI. Volume will be distributed and the idea
> is to have RHCS migrate iSCSI LUN & gluster daemon to a single node in
> case of failure of one node. Due to a limited storage we cannot go for
> replicated volume.
>
> We have a similar setup with earlier version of Gluster (3.0) that
> enabled specification of IP address that each gluster daemon should use.
> This enabled usage of virtual IPs and integration with RHCS with service
> resource agent.
>
> Questions are:
> 1. Is there any documentation on how to use OCF resource agents? I
> checked the source code (glusterd and volume) and the metadata, but
> couldn't figure out how to get two glusterd instances on a same node.
>
> 2. Is it possible to have multiple glusterd instances on a same machine
> using virtual IP?
>
> Thanks in advance


-- 
Emir Imamagic
SRCE - University of Zagreb University Computing Centre, www.srce.unizg.hr
Emir.Imamagic at srce.hr, tel: +385 1 616 5809, fax: +385 1 616 5559