Skip to contentSkip to conversation

Multi-path IO in 2.5/2.6 ?

RFC Lars Marowsky-Bree 3 participants 3 messages Sep 9, 2002 – Sep 10, 2002

lore
Share & tools
lore
#0 lore raw

Morning everyone,

I hope people are waking up by now ;-)

So, what is the take on "multi-path IO" (in particular, storage) in 2.5/2.6?

Right now, we have md multipathing in 2.4 (+ an enhancement to that one by Jens Axboe and myself, which however was ignored on l-k ;-), an enhancement to LVM1 and various hardware-specific and thus obviously wrong approaches.

I am looking at what to do for 2.5. I have considered porting the small changes from 2.4 to md 2.5. The LVM1 changes are probably and out gone, as LVM1 doesn't work still.

I noticed that EVMS duplicates the entire md layer internally (great way to code, really!), so that might also require changing if I update the md code.

Or can the LVM2 device-mapper be used to do that more cleanly?

I wonder whether anyone has given this some thought already.

Sincerely,
Lars Marowsky-Brée <[email protected]>

Signature · 2 lines
Immortality is an adequate definition of high availability for me.
	--- Gregory F. Pfister
#1 lore raw
Quoted 3 lines · Lars Marowsky-Bree

Or can the LVM2 device-mapper be used to do that more cleanly?

I wonder whether anyone has given this some thought already.

The md layer code can already do the job fine - but it does need to get to the point that the block layer provides better error information upstream so it can make better decisions.

LVM2 is a nice clean remapper, so it should sit on top of the md or other failover mappers easily enough. You can probably do failover by updating map tables too.

Its nice clean code unlike EVMS, and doesnt duplicate half the kernel so its easier to hack on

#2 lore raw
In reply to Lars Marowsky-Bree
Quoted 10 lines · Lars Marowsky-Bree

Morning everyone,

I hope people are waking up by now ;-)

So, what is the take on "multi-path IO" (in particular, storage) in 2.5/2.6?

Right now, we have md multipathing in 2.4 (+ an enhancement to that one by Jens Axboe and myself, which however was ignored on l-k ;-), an enhancement to LVM1 and various hardware-specific and thus obviously wrong approaches.

Quoted 8 lines

I am looking at what to do for 2.5. I have considered porting the small changes from 2.4 to md 2.5. The LVM1 changes are probably and out gone, as LVM1 doesn't work still.

I noticed that EVMS duplicates the entire md layer internally (great way to code, really!), so that might also require changing if I update the md code.

Or can the LVM2 device-mapper be used to do that more cleanly?

We have a multi-path target for device-mapper planned for later this year.

This will be a multi-path addon to the generic mapping service(s) device-mapper already provides which can multi-path access to any arbitrary given block device, not just logical volumes.

Quoted 2 lines

I wonder whether anyone has given this some thought already.

We did ;)

Quoted 14 lines

Sincerely,
Lars Marowsky-Brée <[email protected]>

Signature · 8 lines
Immortality is an adequate definition of high availability for me.
	--- Gregory F. Pfister

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
Signature · 12 lines
Regards,
Heinz    -- The LVM Guy --

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-

Heinz Mauelshagen                                 Sistina Software Inc.
Senior Consultant/Developer                       Am Sonnenhang 11
                                                  56242 Marienrachdorf
                                                  Germany
[email protected]                           +49 2626 141200
                                                       FAX 924446
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-