[OpenAFS-devel] Re: idle dead timeout processing in clients

[Andrew Deason]

On Wed, 30 Nov 2011 12:53:03 -0500
Jeffrey Altman <jaltman@your-file-system.com> wrote:

> For 1.6.1 we propose that:
> 
> 1.  Since client side idle peer detection is inherently broken that
>     it be disabled entirely on Unix clients.

So, based on the rest of the thread and other discussions, what is the
current plan for what we're doing with this for unix 1.6.1? While I
agree with the "idle dead is broken/evil/stupid (except for the
exceptions)" points expressed, I would think that reverting the
client-side behavior to something 1.4-ish would be good enough for 1.6.1
without introducing additional complexity or large changes (or 1.4
regressions). I think that this is about as simple as removing the
rx_SetServerConnIdleDeadErr call in afs/afs_conn.c, but since this was
introduced in that mtu ping commit, I'm not sure if this is required for
the mtu stuff.

I think the only commits I've seen for this appear to be part of the
larger "fix idle dead" effort, but it's entirely possible I missed
something. I was assuming we'd do something small and quick for 1.6.1,
with the larger fixes going on master and possibly later on 1.6.x.
Yes/no?

-- 
Andrew Deason
adeason@sinenomine.net