[ClusterLabs] iSCSI on ZFS on DRBD

Jason A Ramsey jason at eramsey.org
Tue Nov 22 18:28:13 CET 2016


The way that Pacemaker interacts with services is using resource agents. These resource agents are bash scripts that you can modify to your heart’s content to do the things you want to do. Having worked with the ocf:heartbeat:iSCSITarget and ocf:heartbeat:iSCSILogicalUnit quite a lot in the last several months, I can tell you that they only support iet, tgt, lio, and lio-t implementations of the standard out of the box. I’m sure you could make modifications to them (I have to support very specific use cases on my NAS cluster, and I’m definitely not a code monkey) as needed. Just take a peak at the resource agent files relevant to what you’re doing and go from there (on my systems they are located at /usr/lib/ocf/resource.d). Good luck!

--

[ jR ]

  there is no path to greatness; greatness is the path

From: Mark Adams <mark at openvs.co.uk>
Reply-To: Cluster Labs - All topics related to open-source clustering welcomed <Users at clusterlabs.org>
Date: Tuesday, November 22, 2016 at 11:59 AM
To: "Users at clusterlabs.org" <Users at clusterlabs.org>
Subject: [ClusterLabs] iSCSI on ZFS on DRBD

Hi All,

Looking for some opinions on this, if anyone has any. I'm looking at this solution to be for proxmox vm nodes using zfsoniscsi.

Just as back round for people that haven't looked at proxmox before it logs on to the iscsi server via ssh and creates a zfs dataset then adds iscsi config to /etc/ietd.conf so that dataset is available as a LUN. This works fine when you've got a single iscsi host, but I haven't figured out a way to use it with pacemaker/corosync.

Is there any way to have ISCSILogicalUnit read it's luns from a config file instead of specifying each one in the cluster config? or is there any other resource agents that might be more suitable for this job? I could write my own "watcher" script I guess, but does anyone think this is a dangerous idea?

Is the only sensible thing really to make proxmox zfsonlinux pacemaker/corosync "aware" so that it's scripts can create the luns through pcs instead of adding the config to ietd.conf?

Is anyone using zfs/iscsi/drbd in some other configuration and had success?

Looking forward to all ideas!

Regards,
Mark
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://clusterlabs.org/pipermail/users/attachments/20161122/6b86e639/attachment.html>


More information about the Users mailing list