Hi,

On Wed, May 05, 2010 at 08:06:39AM -0600, Greg Woods wrote:
> On Wed, 2010-05-05 at 13:29 +0200, Dejan Muhamedagic wrote:
> 
> > If these servers have a lights-out device and the power
> > distribution is fairly reliable, that could be an alternative for
> > fencing.
> 
> They do have an IPMI device and it does work. I am trying to insulate
> against a failure of the NIC or cable by having a second stonith device.
> 
> The cluster I have now is primarily for testing, but eventually we will
> be implementing critical services (e.g. DNS, e-mail, DHCP, and
> authentication) in virtual machines running on a cluster like this one,
> so part of the testing process is to learn what can and can't be done
> and where the potential gotchas are. I have discovered that if I
> simulate a cable failure by removing it, bad things happen because
> stonith cannot succeed. I would not want my DNS system to be vulnerable
> to a single cable failing, so I am looking for ways to guard against it.

rackpdu also works over the network. The lights-out device has
its own network interface. Presumably both should be connected to
some management network. Where's the difference?

> A complete power outage on one of the nodes also results in bad things
> when using IPMI. Again stonith cannot succeed and so the remaining
> server will not take over the resources. Yes, these are dual power
> supply servers so it is unlikely that something would happen that causes
> only one of the servers to completely lose power other than human error
> (possibly a motherboard failure as well?)

That's an interesting question. Perhaps the server vendor can
tell.

> but I am still looking to
> determine if there is a way to guard against this. Right now I have a
> "meatware" stonith device set up so that I can at least log in remotely
> and manually force the remaining server to take over, but I am looking
> for something more automatic. It would be nice to avoid those 3AM phone
> calls )-:

You can have more than one stonith resource and they'll be tried
in a round-robin fashion until one succeeds.

> I may take a shot at modifying the external/rackpdu stonith plugin at
> some point. We can't be the only ones in the world using dual power
> supply servers. I'll probably start by unplugging one of the power
> supplies on each server and making sure I understand how to use the
> plugin in single-outlet mode, then try doing the modifications to
> support dual outlets.

The plugin picks the outlet in one of the two ways:

- from the device itself, provided that the outlet name matches
  the node name
- from the external file which specifies the mapping

I can help with modifying the plugin if you'll do the testing.

Thanks,

Dejan

> --Greg
> 
> 
> _______________________________________________
> 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

Reply via email to