Re: [Scst-devel] More on concurrent writes and multipathing

[Gennadiy N. <pa...@gm...>]

On Wed, Apr 15, 2009 at 2:19 AM, Bart Van Assche
<bar...@gm...>wrote:

> On Tue, Apr 14, 2009 at 10:02 PM, Gennadiy Nerubayev <pa...@gm...>
> wrote:
> > I've just observed DRBD complaining again about concurrent local writes
> with
> > a single SRP initiator using a single path to a single target. At first
> I've
> > thought that it was due to scst multithreading, but scst_threads=1 still
> has
> > the same effect. Example:
> >
> > Apr 14 13:35:11 srpt2 kernel: drbd0: scsi_tgt0[12503] Concurrent local
> write
> > detected! [DISCARD L] new: 144072784s +8192; pending: 144072784s +8192
>
> This looks like a good question for the DRBD people: I'm not sure
> whether this is a DRBD feature or a DRBD bug. More information about
> the DRBD mailing lists can be found here:
> http://www.drbd.org/home/mailinglists/.


As I suspected, this is apparently a feature:

http://lists.linbit.com/pipermail/drbd-user/2009-April/011875.html

I played with the settings a little bit more, and after some additional
trial and error, I discovered that using NV_CACHE mode no longer triggers
concurrent write warnings. As soon as I change back to BLOCKIO, the warnings
come back. I'm a little bit apprehensive abount using NV_CACHE however (for
reasons stated in the readme), so I'd still like to figure out why one mode
causes concurrent warnings and one doesn't.

Thanks,

-Gennadiy