All of lore.kernel.org
 help / color / mirror / Atom feed
* Use device mapper to redirect to cache server?
@ 2008-01-08  9:30 Ph. Marek
  2008-01-08 19:30 ` malahal
  0 siblings, 1 reply; 2+ messages in thread
From: Ph. Marek @ 2008-01-08  9:30 UTC (permalink / raw)
  To: linux-kernel; +Cc: dm-devel

Hello everybody!


I'd like to mirror write accesses to a block-device to two devices, get all 
all *read* accesses from one of these, and if it fails, to fallback to the 
other device.


The usage would be to use some kind of NAS box as persistent network block 
device, with some other machine with a lot of RAM as cache server; the cache 
server would get asked for data, and if it's not available, the request 
should go to the NAS directly.

If I read the multipath documentation correctly (at 
http://christophe.varoqui.free.fr/multipath.html):
	For completion, here is an example of a pure failover target definition for
 	the same LU :
	...
multipath should do part of what I want - use another device if the first one 
fails.
But that doesn't account for pushing writes to both; and how is failing 
defined for NBD? I don't see some configurable timeout.


Thank you for all ideas/suggestions! [[ but please CC me ]]


Regards,

Phil

^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: Use device mapper to redirect to cache server?
  2008-01-08  9:30 Use device mapper to redirect to cache server? Ph. Marek
@ 2008-01-08 19:30 ` malahal
  0 siblings, 0 replies; 2+ messages in thread
From: malahal @ 2008-01-08 19:30 UTC (permalink / raw)
  To: dm-devel; +Cc: philipp.marek

Ph. Marek [philipp.marek@bmlv.gv.at] wrote:
> Hello everybody!
> 
> 
> I'd like to mirror write accesses to a block-device to two devices, get all 
> all *read* accesses from one of these, and if it fails, to fallback to the 
> other device.

This is exactly what 'mirror' target does (dm-raid1.c) without
read-balancing. But read on...

> The usage would be to use some kind of NAS box as persistent network block 
> device, with some other machine with a lot of RAM as cache server; the cache 
> server would get asked for data, and if it's not available, the request 
> should go to the NAS directly.

I think, you are better of using the cache server as some kind of NFS
server as 'mirror' target's failure handling is heavy-weight.

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2008-01-08 19:30 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2008-01-08  9:30 Use device mapper to redirect to cache server? Ph. Marek
2008-01-08 19:30 ` malahal

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.