On Thu, Aug 5, 2010 at 12:12 PM, Gary Sedgwick <[email protected]> wrote: > Hi, > > I have two identical basic Dell Poweredge R300 servers - when I say > "basic", I mean no DRAC card, no redundant power supplies, etc. Each has > a smaller disk (120GB) and larger disk (1TB), as well as 2 NICs. I'm > planning to run (production) Apache/Tomcat/MySQL/DNS/e-mail services using > a BlueOnyx VM. > > I've been experimenting with Proxmox (OpenVZ) and DRBD (in > primary/secondary due to BlueOnyx requiring an ext3 filesystem) to > replicate the larger 1TB disk (using the 2nd NIC on each server, connected > with a crossover cable), both of which seem to successfully do what I > want. I've now been looking into how I could use Pacemaker for > HA/failover (i.e. automatic takeover of IP and services in case of failure > of a server)... > > The conclusion I'm rapidly coming to is that I'd probably be better off > (less risk/complexity) configuring the machines to notify me in case of > error, and to then manually shutdown and migrate VMs on the problem > server. Especially as I don't (or at least don't believe I) have suitable > devices for fencing/STONITH, etc., and adding hardware (DRAC cards, > redundant power supplies etc.) isn't an option. So I have some questions > that I would very much appreciate people lending their experience to: > > 1) Is it worth persevering with setting up Pacemaker/HA? Am I able to set > up a fairly simple (low risk) configuration that could provide automatic > failover of services (not necessarily in all failure scenarios), even if > not removing every single point of failure?
I believe so. > 2) If so, is it possible to have a fairly "safe" configuration that would > migrate services in common failure situations (in my experience, disk > errors!) but that would not have the possibility of split brain etc.? Well... you're using a crossover cable, so the risk of split-brain is lower than normal. But even with stonith you can't remove the possibility of split-brain - stonith "just" gives you a sane way to keep hosting services when it happens. Perhaps no-quorum-policy=freeze could be of use in your scenario. > 3) If not, it is easy to configure Pacemaker to simply alert a sysadmin of > problems? on-fail=block > > Many thanks for any advice. > > Gary > _______________________________________________ > Linux-HA mailing list > [email protected] > http://lists.linux-ha.org/mailman/listinfo/linux-ha > See also: http://linux-ha.org/ReportingProblems > _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
