[dm-devel] [PATCH 2/2] multipathd: add recheck_wwid_time option to verify the path wwid

[Martin Wilck]

On Thu, 2021-02-18 at 11:22 +0800, Chongyun Wu wrote:
> > 
> > I wasn't aware of multiple affected users. I saw Chongyun's post
> > and it
> > looked to me as if this had happend once, likely in his
> > organization.
> > It wasn't even clear to me whether production data were affected.
> > 
> Had customers have made incorrect storage back-end operations during
> the 
> expansion process, which has caused the destruction of customer data,
> which 
> cannot be repaired using any tools. And I reproduce it in our lab
> too.
> Customers definitely hope that multipath-tools can probide protection
> in
> this case, becuase if it is important custmer data storage in the
> LUN, the
> loss will be great.

I understand that. My concern was that it's still just quite a weak
protection, because it depends on the check time interval being short
enough to catch the change. So in a way, this pretends to offer safety
that doesn't truly exist. If the user assigns the wrong WWID 30s after
taking the LUN offline, all fine; if she is quicker (or has the process
automated / scripted), data corruption still occurs.

And of course, there are general limits to which extent software can
avoid user mistakes.

However, I trust your and Ben's judegement that by checking this in a
suitably short time span, the issues your customers have had would have
been avoided, and I'm not going to oppose the patch on these grounds.

Martin

-- 
Dr. Martin Wilck <mwilck at suse.com>, Tel. +49 (0)911 74053 2107
SUSE Software Solutions Germany GmbH
HRB 36809, AG Nürnberg GF: Felix Imendörffer