[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [dm-devel] Multipath Questions

On Wed, Jan 14, 2004 at 01:36:49PM -0600, Kevin Corry wrote:
> Hi,
> I just had a couple questions about the overall design and direction of the DM 
> multipath module.
> 1. Priority Groups.
> DM-multipath allows paths to be placed in groups, with each group assigned a 
> priority (0 to 1024). What is the intended purpose for these groups? In the 
> current code, if a device has multiple groups, only the lowest-priority group 
> is used. I'm assuming this will change, but I just wanted to get an idea of 
> how the group priority will affect path selection.


low numbers imply high priority (ie, group with prio 1 is used before
group with prio 3).

What this actually comes down to is, that io gets dispatched to the group
with the highest priority (lowest #) until _all_ its paths become faulty.

One point to take into consideration later would be:
do we want to hang on a higher priority group with just 1 path operational,
_while_ we've got lower priority ones with multiple paths ?

> 2. Multipath Daemon
> The daemon's job is to resubmit failed I/Os to a different path, and to 
> periodically test each of the paths to see if they've failed or recovered 
> from failure. However, currently the daemon only wakes up on the error path 
> for normal I/Os. This means a path could be marked faulty, but never 
> reactivated if the path gets fixed.
> Obviously the daemon needs a way to wake-up periodically so it can run it's 
> test I/Os. The comments in dm-daemon.h say the worker function should return 
> a "hint" as to when they should be woken up, but currently this return value 
> is unused. I'm trying to look into how to implement this functionality in the 
> dm-daemon code, but I figured I'd ask here and see if anyone has already come 
> up with a plan for this. Once we get the multipath daemon to do its periodic 
> path-testing, I think we'll have a working prototype.

Joe, is that still the case (it is supposed to wake up regularly and test io
on failed paths as Kevin suggests) ?

> Thanks!
> -- 
> Kevin Corry
> kevcorry us ibm com
> http://evms.sourceforge.net/
> _______________________________________________
> dm-devel mailing list
> dm-devel sistina com
> http://lists.sistina.com/mailman/listinfo/dm-devel


Heinz    -- The LVM Guy --

*** Software bugs are stupid.
    Nevertheless it needs not so stupid people to solve them ***


Heinz Mauelshagen                                 Sistina Software Inc.
Senior Consultant/Developer                       Am Sonnenhang 11
                                                  56242 Marienrachdorf
Mauelshagen Sistina com                           +49 2626 141200
                                                       FAX 924446

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]