[Linux-cluster] How to monitor secondary link?

[Jakov Sosic]

Hi.

I have the following problem (RHEL v5.3) with linux cluster, 3 node
cluster configuration with qdisk.

I have two interfaces - bond0 and bond1. Bond0 is used for cluster
communication (heartbeat) and virtual ip addresses for services plus
it's the interface that's bridged for Xen machines.

Bond1 is used only for accessing iSCSI NAS storage. NAS is configured
with CLVM.

Problem arises when link to NAS is down (ifdown bond1).
Cluster doesn't relocate Xen's to the other nodes, so it's obvious that
cluster does not monitor bond1 link.

I have three ideas how to solve this, but I'm not too happy about
neither of them.

1. heuristic pings in qdisk configuration.

I'm not too happy about this one because I would had to add all three
nodes to heuristics, and tune the ping ratio... And maybe when there
would be bursts and those links too occupied, some ping packets could
get lost and get me in trouble.

2. move heartbeat over to bond1

That would get me in new trouble, because I'm doing the fencing through
bond0, so... :-/ And also, again there's a potentially problem when
services start to really push the storages and traffic to storage goes
up.


3. adding 3 virtual IP addresses on bond1 network, and adding them to
some services, each service running on one of the nodes.

I don't like this because it's a "hack" and not a real solution :-/ But
it seems to be the best choice for the moment... Problem is I also
didn't test this one, and this might resolve in IP relocating, and
Xen's staying in place :-/


Any ideas?



-- 
|    Jakov Sosic    |    ICQ: 28410271    |   PGP: 0x965CAE2D   |
=================================================================
| start fighting cancer -> http://www.worldcommunitygrid.org/   |