[dm-devel] multibus / failover and EMC CX600

[Hannes Reinecke]

On Thu, Oct 18, 2007 at 08:55:52AM +0200, Gerald Nowitzky wrote:
> Hannes,
> 
> so is this behavior by design? In this case, patching in the scsi subsystem won't
> help to much, will it?
>
No, not really.
 
> Manually updating the multipath information - well, yes, that will work I guess,
> but in the end that should work without manual intervention. Thus, I'd need to have
> a job checking the multipath table if there are stale devices, and, if there are,
> rerun multipath to get them out. Not exactly smooth, is it?
>
Why, we do have interns for this kind of work :-)

No, really: Normally this should be done by multipathd / udev. With the mainline
multipathd it reads from the kernel netlink socket and will get a uevent if
a device is removed. And it should update the tables accordingly.

There have been problems with device-mapper itself (older versions required to
flush all outstanding I/O before the table could be modified), but that should
be resolved by now.

So maybe have multipathd running in verbose mode (ie -v 6 or somesuch) and see
what's going on. Especially why it isn't updating the device-mapper tables.

BTW, which distribution are you running? multipath & udev seem to be a tricky
area for most.

Not ours, of course :-)

Cheers,

Hannes
-- 
Dr. Hannes Reinecke		      zSeries & Storage
hare at suse.de			      +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: Markus Rex, HRB 16746 (AG Nürnberg)