[dm-devel] [PATCH 1/3] libmultipath: replace PATH_TIMEOUT with PATH_DOWN

[Hannes Reinecke]

On 10/04/2014 02:11 AM, Benjamin Marzinski wrote:
> The way the code works, PATH_TIMEOUT is treated mostly like PATH_UP or
> PATH_GHOST by check_path.  If the the path was previously failed,  it will
> even reinstate the path. It will also trigger prio refreshing. It seems
> that PATH_TIMEOUT should be at least as serious as PATH_PENDING, but the
> way the code works, it's not.  In pathinfo, PATH_TIMEOUT gets changed
> directly to PATH_DOWN, which makes sense.  But assuming that's the correct
> thing to do, why have PATH_TIMEOUT at all?
> 
Because a timeout is different from a normal path down.
Timeout means the tur checker is stuck somehow.
And we currently have no real means of resetting it (aio_cancel
doesn't really abort the I/O, is just short-circuit the callback).

So the intention of this patch was that we want to get notified if
a TUR timeout occurs, as this might lead to other subsequent errors.

> The only thing that it does that seems helpful is that when you print out
> the path, instead of it saying that the path is down, it says that the
> path has timed out.  But if we are going to treat is like the path is
> down, then I don't see this being too helpful.  And the way we are treating
> PATH_TIMEOUT right now is definitely not right.
> 
See above. I really would like to be notified for PATH_TIMEOUT
scenarios ...

Cheers,

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