|
Or maybe I'm not thinking of the simple solution...
I've set up a LVS-DR cluster on Fedora Core 3
The primary purpose will be to run scripts that are called via ssh to the
virtual IP.
I can't use cron on one of the cluster members to call the scripts -- since
if I hit the VIP, the machine running the cron will always process the job.
I'm currently running jobs, that are being distributed OK, by cron'ing them
from one of my non-clustered Linux boxes as well as the MKS Toolkit
Scheduler from one of my NT servers.
Neither of those is a proper solution, since they would represent a single
point of failure.
So what I am thinking is configuring two additional older PCs from our pile
of miscellaneous old machines and clustering them together with Keepalived
just to run cron.
Something along the lines of each normal runs half the jobs, but if
Keepalived detects a failover, it trips a script that writes a crontab that
has both sets of jobs in it, and restarts the cron program.
Questions:
1) Can Keepalived run on a machine that doesn't have LVS installed? (I
don't need a virtual address or anything, just the cron!)
2) Is this way more work than it should and there is a simpler, more
elegant solution out there?
|
