Hi all,
I've been thinking about configuring clusters, particularly in a
virtualised environment. The solutions we have for this are not
especially good I don't think.
We currently have two loaders:
xmlconfig is the default and reads from /etc/cluster/cluster.conf (by
default)
ldapconfig reads from an LDAP server
Both of these have their uses of course. But, especially in a
virtualised environment, they are not particularly handy. LDAP is
cumbersome and complicated to configure and we have no real tools to
update it directly. Reading from local files seems a bit primitive in a
virtualised system where you have at least one host system that is
guaranteed to be available because it is actually running the guests
themselves.
Yes, we have ccs_sync that will synchronise the configuration across
nodes, and that is integrated into 'cman_tool version' now, but it still
seems cumbersome to me and doesn't work properly if a cluster node is
down when the configuration is changed.
What I think would be nice would be to update the configuration on a
host node (dom0 in Xen parlance) and have the other nodes automatically
pick it up from that one place.
One suggested way of doing this is to host the cluster.conf file on an
HTTP server on the host. The VMs can then simply GET that file and run
it through xmlconfig to load it into corosync. This seems simple and
easy to me. Most people know how to set up an http server (even if they
don't know how to secure one!)
The intention is to make this as easy as possible, so that the admin
staff simply supply a URL and corosync does the right thing to fetch it
when needed (at boot up and re-configure time). This magic is brought to
you courtesy of pluggable configuration modules in corosync :)
Comments? Better ideas? Or am I just barking ?
chrissie
--
Linux-cluster mailing list
Linux-cluster@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/linux-cluster
