Re: [Lsf-pc] [LSF/MM/BPF TOPIC] FS Maintainers Don't Scale

From: "Darrick J. Wong" <darrick.wong@oracle.com>
To: Dave Chinner <david@fromorbit.com>
Cc: Matthew Wilcox <willy@infradead.org>,
	Brian Foster <bfoster@redhat.com>,
	Allison Collins <allison.henderson@oracle.com>,
	Amir Goldstein <amir73il@gmail.com>,
	lsf-pc@lists.linux-foundation.org,
	linux-fsdevel <linux-fsdevel@vger.kernel.org>,
	xfs <linux-xfs@vger.kernel.org>, Eryu Guan <guaneryu@gmail.com>,
	Eric Sandeen <sandeen@redhat.com>
Subject: Re: [Lsf-pc] [LSF/MM/BPF TOPIC] FS Maintainers Don't Scale
Date: Tue, 18 Feb 2020 16:29:16 -0800	[thread overview]
Message-ID: <20200219002916.GB9506@magnolia> (raw)
In-Reply-To: <20200216215556.GZ10776@dread.disaster.area>

On Mon, Feb 17, 2020 at 08:55:56AM +1100, Dave Chinner wrote:
> On Thu, Feb 13, 2020 at 07:46:32AM -0800, Matthew Wilcox wrote:
> > On Thu, Feb 13, 2020 at 10:11:00AM -0500, Brian Foster wrote:
> > > With regard to the burnout thing, ISTM the core functionality of the
> > > maintainer is to maintain the integrity of the subtree. That involves
> > > things like enforcing development process (i.e., requiring r-b tags on
> > > all patches to merge), but not necessarily being obligated to resolve
> > > conflicts or to review every patch that comes through in detail, or
> > > guarantee that everything sent to the list makes it into the next
> > > release, etc. If certain things aren't being reviewed in time or making
> > > progress and that somehow results in additional pressure on the
> > > maintainer, ISTM that something is wrong there..?
> > > 
> > > On a more logistical note, didn't we (XFS) discuss the idea of a
> > > rotating maintainership at one point? I know Dave had dealt with burnout
> > > after doing this job for quite some time, Darrick stepped in and it
> > > sounds like he is now feeling it as well (and maybe Eric, having to hold
> > > down the xfsprogs fort). I'm not maintainer nor do I really want to be,
> > > but I'd be willing to contribute to maintainer like duties on a limited
> > > basis if there's a need. For example, if we had a per-release rotation
> > > of 3+ people willing to contribute, perhaps that could spread the pain
> > > around sufficiently..? Just a thought, and of course not being a
> > > maintainer I have no idea how realistic something like that might be..
> > 
> > Not being an XFS person, I don't want to impose anything, but have
> > you read/seen Dan Vetter's talk on this subject?
> > https://blog.ffwll.ch/2017/01/maintainers-dont-scale.html (plenty of
> > links to follow, including particularly https://lwn.net/Articles/705228/ )
> 
> Yes, and I've talked to Dan in great detail about it over several
> past LCAs... :)
> 
> > It seems like the XFS development community might benefit from a
> > group maintainer model.
> 
> Yes, it may well do. The problem is the pool of XFS developers is
> *much smaller* than the graphics subsystem and so a "group
> maintainership" if kinda what we do already. I mean, I do have
> commit rights to the XFS trees kernel.org, even though I'm not a
> maintainer. I'm essentially the backup at this point - if someone
> needs to take time off, I'll take over.
> 
> I think the biggest barrier to maintaining the XFS tree is the
> amount of integration testing that the maintainer does on the merged
> tree.  That's non-trivial, especially with how long it takes to run
> fstests these days. If you're not set up to run QA 24x7 across a
> bunch of different configs, then you need to get that into place
> before being able to do the job of maintaining the XFS kernel
> tree...
> 
> If everyone had that capability and resources at their disposal, then
> rotating the tree maintenance responsibilities would be much
> easier...

I kinda wish the LF or someone would open such a program to the kernel
maintainers.  I never liked that old maxim, "The maintainer is [the
stuckee] with the most testing resources" -- there shouldn't really have
to be a djwong cloud and a dchinner cloud. :/

--D

> Cheers,
> 
> Dave.
> -- 
> Dave Chinner
> david@fromorbit.com