All of lore.kernel.org
 help / color / mirror / Atom feed
* Question about Multibus+Failover
@ 2005-12-06 20:37 Josef Whiter
  2005-12-06 21:15 ` Christophe Varoqui
  0 siblings, 1 reply; 4+ messages in thread
From: Josef Whiter @ 2005-12-06 20:37 UTC (permalink / raw)
  To: device-mapper development

Hello,

I've had several requests about loadbalancing coupled with failover for 
dm multipathing.  I know with failover you get just failover, and 
multibus you get just load balancing, but I was wondering if there was a 
hybrid option for load balancing that would handle a failed path 
gracefully.  Thanks much,

Josef

-- 
Josef Whiter, RHCE
Global Support Services
Red Hat, Inc.
919-754-3700 x44429

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: Question about Multibus+Failover
  2005-12-06 21:15 ` Christophe Varoqui
@ 2005-12-06 21:10   ` Josef Whiter
  2005-12-06 21:41     ` Christophe Varoqui
  0 siblings, 1 reply; 4+ messages in thread
From: Josef Whiter @ 2005-12-06 21:10 UTC (permalink / raw)
  To: device-mapper development

Christophe Varoqui wrote:

>On mar, 2005-12-06 at 15:37 -0500, Josef Whiter wrote:
>  
>
>>Hello,
>>
>>I've had several requests about loadbalancing coupled with failover for 
>>dm multipathing.  I know with failover you get just failover, and 
>>multibus you get just load balancing, but I was wondering if there was a 
>>hybrid option for load balancing that would handle a failed path 
>>gracefully.  Thanks much,
>>
>>    
>>
>Failed paths are handled quite gracefully in all grouping policies.
>
>That said, other stock grouping policies are at your disposal :
>group_by_serial or group_by_node_name or the special group_by_prio that
>let you do your own grouping policy with callouts (shell scripts or
>whatever).
>
>Please have a look at the documentation available.
>
>Regards,
>cvaroqui
>
>  
>
Hmm, then perhaps I've been seeing errant behavior, b/c everytime 
somebody uses multibus and they fail a path they get a bunch of IO 
errors and the FS is remounted, but if they use failover and pull the 
active connection, there are no errors and the failover is seamless.  I 
shall investigate further on what they are seeing.  Thank you,

Josef

-- 
Josef Whiter, RHCE
Global Support Services
Red Hat, Inc.
919-754-3700 x44429

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: Question about Multibus+Failover
  2005-12-06 20:37 Question about Multibus+Failover Josef Whiter
@ 2005-12-06 21:15 ` Christophe Varoqui
  2005-12-06 21:10   ` Josef Whiter
  0 siblings, 1 reply; 4+ messages in thread
From: Christophe Varoqui @ 2005-12-06 21:15 UTC (permalink / raw)
  To: dm-devel

On mar, 2005-12-06 at 15:37 -0500, Josef Whiter wrote:
> Hello,
> 
> I've had several requests about loadbalancing coupled with failover for 
> dm multipathing.  I know with failover you get just failover, and 
> multibus you get just load balancing, but I was wondering if there was a 
> hybrid option for load balancing that would handle a failed path 
> gracefully.  Thanks much,
> 
Failed paths are handled quite gracefully in all grouping policies.

That said, other stock grouping policies are at your disposal :
group_by_serial or group_by_node_name or the special group_by_prio that
let you do your own grouping policy with callouts (shell scripts or
whatever).

Please have a look at the documentation available.

Regards,
cvaroqui

^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: Question about Multibus+Failover
  2005-12-06 21:10   ` Josef Whiter
@ 2005-12-06 21:41     ` Christophe Varoqui
  0 siblings, 0 replies; 4+ messages in thread
From: Christophe Varoqui @ 2005-12-06 21:41 UTC (permalink / raw)
  To: device-mapper development


> >  
> >
> Hmm, then perhaps I've been seeing errant behavior, b/c everytime 
> somebody uses multibus and they fail a path they get a bunch of IO 
> errors and the FS is remounted, but if they use failover and pull the 
> active connection, there are no errors and the failover is seamless.  I 
> shall investigate further on what they are seeing.  Thank you,
> 
May be because they use hardware that needs a hardware handler ?
The hwh trigger needs different path group, thus not compatible with
multibus.

Regards,
cvaroqui

^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2005-12-06 21:41 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2005-12-06 20:37 Question about Multibus+Failover Josef Whiter
2005-12-06 21:15 ` Christophe Varoqui
2005-12-06 21:10   ` Josef Whiter
2005-12-06 21:41     ` Christophe Varoqui

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.