Your browser doesn't support the features required by impress.js, so you are presented with a simplified version of this presentation.

For the best experience please use the latest Chrome, Safari or Firefox browser.

GlusterFS in High Availability Clusters

Integration with the Pacemaker cluster stack

Gluster Workshop, LinuxCon Europe 2012

Barcelona, Catalunya, España

Nov 8, 2012

In the next 45 minutes, I will give you an overview of GlusterFS/Pacemaker integration.

We'll break this down into 3 parts.

1

What's Pacemaker, really?

2

How do I use GlusterFS storage for Pacemaker?

3

How do I manage GlusterFS storage with Pacemaker?

1

Pacemaker is the state of the art high availability stack for Linux.

It is built on the Corosync cluster messaging subsystem.

It is entirely application agnostic.

Application management is through resource agents.

70

It is the default cluster stack on Debian, SUSE, Ubuntu, and CentOS.

Pacemaker is a Technology Preview in RHEL 6, and will be the default HA stack in RHEL 7.

Here is a simple 3 node Pacemaker cluster running on Debian.

alice

bob

charlie

2

Pacemaker is storage agnostic.

One storage option for Pacemaker: NAS

As in CIFS or NFS.

Simple and easy, no locking issues.

But: very limited scaleout capability

Another: single-instance shared storage.

Requires GFS, OCFS2 or similar.

Another: DRBD.

Limited scalability, limited multi-master support.

Pacemaker allows us to easily use GlusterFS mounts for storage.

We can use the general-purpose ocf:heartbeat:Filesystem resource agent.

alice

bob

charlie

3

In 3.4, GlusterFS will ship with Pacemaker resource agents to manage both glusterd and individual volumes.

They allow us to auto-recover GlusterFS services on systems where neither upstart nor systemd are available.

ocf:glusterfs:glusterd manages glusterd on all GlusterFS server nodes.

ocf:glusterfs:volume manages individual bricks.

alice

bob

charlie

And of course we can also combine everything.

What we could be doing in the near future:

100% open-source cross-site high availability with automatic failover.

Can you hear a manager squeal?

Pacemaker can already do this.

GlusterFS can't (yet).

Thanks to:

Bartek Szopka @bartaz for impress.js

Markus Gutschke for shellinabox (and his recipe collection!)

This talk:

http://www.hastexo.com/lceu2012-glusterfs

https://github.com/fghaas/lceu2012