Minneapolis Cluster Summit 2004
Presentations
The first ever Mineapolis Cluster Summit took place in July this
summer. This was initially planned to be an informal workshop to
bring GFS developers face-to-face and kick off community development of
the GFS cluster filesystem suite, recently acquired and liberated by
Red Hat. Our little workshop grew into a meeting of over 50
prominent cluster developers from all over the world.
The summit workshop began on Thursday with Red Hat presentations of the
(former) Sistina
GFS cluster filesystem and cluster infrastructure, continuing to Friday
morning. On friday afternoon we moved on to presentations
from other community cluster projects. We concluded with a round
table discussion on adapting GFS cluster infrastructure components for
use by other projects, and how we will prepare for mainline kernel
submission. It is fair to
say
that much solid work was revealed to all participants, in many cases,
for the first time.
- Summary
by John Cherry
- Red Hat Cluster
Infrastructure Overview
- GULM - User Space
client/server cluster membership and lock service
- CMAN - Kernel cluster
membership
- DLM - Kernel
distributed lock manager
- Magma - User level
Cluster and Lock manager transparent library
- GFS - Evolution and
changes between GFS 4.2 to today
- Global Network Block
Device (GNBD)
- CLVM - Architecture
and extensions of LVM2
- Cluster Block Devices
(Snapshot, Mirror)
- Cluster resource
management
- OpenSSI
- SCRAT + Resource
Agent API
- Linux HA
- Clustered samba
- Related Material
Summary by John
Cherry
John Cherry of OSDL wrote a summary for the OSDL mailing lists and
kindly consented have it included here:
The summit was well attended. Besides RedHat/Sistina, other open source
cluster projects were represented (linux-ha, OpenSSI) as well as
proprietary clusters projects and cluster filesystems (i.e. Lustre,
OCFS...).
The meetings were primarly intended to inform the community about the
recently opened RHAT/Sistina clustering components (including clustered
LVM and GFS).
The time was right to consider common clusters components. While we
expected a fair amount of contention at the meetings, it was good to see
a fairly unanimous desire to identify common components that could be
leveraged over the various cluster implementations and to drive these
common components to mainline acceptance. The common cluster components
identified at the summit were...
cman - cluster manager (membership/quorum/heartbeat, recovery
control.
fence - userland daemon which decides which nodes need fencing
dlm - fully distributed, fully symmetrical lock manager
gfs - clustered filesystem
While these common components all have RHAT/Sistina roots, these
components are in the best position for mainline acceptance. As APIs
are defined for these services, other implementations could also be used
(the vfs model).
It was felt that it would be best to leverage an existing mailing list
for discussions regarding the common cluster components, so
linux-cluster will be used for this. To subscribe...
http://www.redhat.com/mailman/listinfo/linux-cluster
The project page has links to the various components and can be found
at:
http://sources.redhat.com/cluster/
The cluster community would like developers to start looking at the code
and testing it in their environments. Documentation, source code,
mailing lists, IRC, and key links can all be found on the project page.
John Cherry
CGL Roadmap Coordinator
Originally posted here.
Red Hat Cluster
Infrastructure Overview
Presented by: Patrick
Caulfield, Red Hat
Slides - Cluster Infrastructure
Slides - CLVMD
GULM - User Space
client/server cluster membership and lock service
Presented by: Mike Tilstra,
Red Hat
Slides - GULM Overview
CMAN - Kernel cluster
membership
Presented by: Patrick
Caulfield, Red Hat
Slides - CMAN
DLM - Kernel
distributed lock manager
Presented by: Patrick
Caulfield, Red Hat
Slides - GDLM
Magma - User level
Cluster and Lock manager transparent library
Presented by: Lon Hohberger,
Red Hat
(bonus feature: Supporting Red Hat's legacy cluster applications)
Slides - Magma Cluster Library
GFS - Evolution and
changes between GFS 4.2 to today
Presented by: Ken Preslan, Red
Hat
Slides - GFS Evolution
Global Network Block
Device (GNBD)
Presented by: Ben Marzinski,
Red Hat
Slides - GNBD Overview
Other useful GNBD documentation:
GFS
6.0 Administrator's Guide (outdated, but most of the information is
still correct)
Amir's
GNBD usage page (Current)
GNBD man pages
Somewhere between 6.0 information and Current. This is available with
the code (cluster/gnbd/man/*).
The documentation will be updated shortly.
CLVM - Architecture
and extensions of LVM2
Presented by: Alasdair Kergon,
Red Hat
Nothing
here yet
Cluster Block Devices
(Snapshot, Mirror)
Presented by: Daniel Phillips,
Red Hat
Slides -
It's all done with Smoke & Mirrors
Slides - Cluster Snapshots
Cluster resource
management
Presented by: Lon Hohberger,
Red Hat
Slides
- Cluster Resources
OpenSSI
Presented by: Bruce Walker, HP
Slides - Linux SSI
Sourceforge
Project Page
openssi.org
SCRAT + Resource
Agent API
Presented by: Lars Marowsky
Bree, Suse/Novell
SCRAT
White Paper
Linux HA
Presented by: Alan Robertson,
IBM
Slides: Evolution
Clustered samba
Presented by: Chris Hertel,
Samba team
Various interesting/nasty issues of event notification, case
translation and database distribution were discussed.
CIFS conference slides - Samba
Clustering with GFS
Samba Implementation - Read the
book
Samba
"Samba moosed be clustered" -- Daniel
"Yes it moosed" -- Chris
Related Material
Presented by: Various
Cluster
Summit Announcement
Cluster
Summit Information and Schedule
Community Cluster Page
Slides from the Linux Cluster
Infrastructure BOF at Ottawa Linux Symposium
OpenSSI
Linux-HA
SRRD
Your Humble Organizers...
Matthew
O'Ke
efe
Daniel
Phillips

And special thanks to
Brian
Stevens

for making it all happen.
The fine print: Material on this page is copyright the individual
authors. Material submitted by Red Hat employees may be freely
copied so long as credit is given to the original author. If
changes are made, the
original author should be notified.