[dm-devel] [PATCH 16/16] multipath: update man pages

[Benjamin Marzinski]

On Wed, Aug 14, 2019 at 09:21:32PM +0000, Martin Wilck wrote:
> On Fri, 2019-08-02 at 11:33 -0500, Benjamin Marzinski wrote:
> > Add documentation for the marginal_pathgroups option and the
> > (un)setmarginal commands.
> > 
> > Signed-off-by: Benjamin Marzinski <bmarzins at redhat.com>
> > ---
> >  multipath/multipath.conf.5 | 34 ++++++++++++++++++++++++++++++----
> >  multipathd/multipathd.8    | 19 +++++++++++++++++++
> >  2 files changed, 49 insertions(+), 4 deletions(-)
> > 
> > --- a/multipathd/multipathd.8
> > +++ b/multipathd/multipathd.8
> > @@ -277,6 +277,25 @@ Remove the persistent reservation key associated
> > with $map from the
> >  \fIreservation_key\fR is set to \fBfile\fR in
> > \fI/etc/multipath.conf\fR.
> >  .
> >  .TP
> > +.B path $path setmarginal
> > +move $path to a marginal pathgroup. The path will remain in the
> > marginal
> > +path group until \fIunsetmarginal\fR is called. This command will
> > only
> > +work if \fImarginal_pathgroups\fR is enabled and there is no Shaky
> > paths
> > +detection method configured (see the multipath.conf man page for
> > details).
> 
> This is counter-intuitive. It's unclear to me why we need these
> commands at all. By nature of "shaky" paths, it doesn't make a lot of
> sense to make these settings (only) manually. I'd like it better if the
> cli commands somehow hooked into the different "shaky" algorithms. E.g.
> for the san_path_err_ algorithm, setting a path to marginal would mean
> artificially setting its failure count to a value above the configured
> threshold. That way, the manual settings could work togehter with the
> automatic detection methods, and could be used for overriding them
> in special cases.

I get that it's weird, but like I mentioned in PATCH 00/16,  The manual
control is for Broadcom's Fiber Channel Transport Daemon, since it
doesn't use the multipathd marginal path detectors, and thus will not
automatically reinstate marginal paths when all other paths have failed.

The way it works right now, the daemon is supposed to have control over
a path's marginal state, without users setting up a different marginal
path algorithm.  I do agree that it would make sense to have a
discussion about automatically restoring paths that have been externally
marked as marginal.

-Ben

> Martin
>