[Linux-cluster] RFC: updating cluster.conf
Sean E. Millichamp
sean at bruenor.org
Tue Jun 24 02:18:02 UTC 2008
On Mon, 2008-06-23 at 11:16 -0500, David Teigland wrote:
> 1. How often do you update cluster.conf? ("Never" would be valuable
> 2. What changes do you make? e.g. add nodes, change fencing settings,
> add or change rgmanager settings.
Currently we rarely update it, only for node adds/removes. We only use
Cluster Suite for fencing and GFS services though. If/when we begin to
use rgmanager I suspect it will get updates frequently (possibly
multiple times a week).
> 3. How do you currently update cluster.conf? Cluster online or offline?
> Manually scp to all nodes? ccs_tool? conga? What do you like and not
> like about the method you use now?
We update cluster.conf in a version controlled repository which we use
cfengine to deploy on the hosts. When cfengine detects a deploy it uses
ccs_tool to update it across the cluster.
I'm not certain if we would use the same mechanism if we were using
rgmanager, but I think we would.
> 4. How would you like to do updates to cluster.conf in the future?
> Conga (graphical management interface)? Command line program that
> updates /etc/cluster/cluster.conf on all cluster nodes?
> Manually scp to all nodes? Other?
I liked Conga, but I got impatient waiting for it. Yanking and pasting
lines in vi ended up being much faster. Also, it didn't fit into our
"every config in version control" concept.
We also place high value on having a single repository where we can diff
all changes made to the environment across all services (if necessary)
over a period of time with the history of who made them. While Conga
could (and may - I don't recall) provide all of this, it is more
convenient to have one central tool for auditing changes (the VC system
logs and diffs) then have to use various auditing tools in various
different GUI interfaces (if all of the services we used even had them).
I think we're probably happy to continue doing updates as we have done
them. Having a validation tool which we could run independently from
any cluster infrastructure for sanity/syntax validation would be nice -
such as having a DTD for cluster.conf. There may be one, but I don't
think I've ever found it.
> 5. Would you like to use an LDAP server? All cluster nodes would read
> cluster.conf info from the server; updates would just be made on
> the server.
Probably not. I can't see what that would really get us given that we
already have a version controlled cfengine setup to help us manage the
history and distribution.
Hope that helps!
More information about the Linux-cluster