[OpenAFS-devel] Problems with shutting down afsd - revisited

[Markus Alt]

Markus Alt wrote:
> 
> Derrick J Brashear wrote:
> >
> > On Sun, 27 May 2001, Markus Alt wrote:
> >
> > > I haven't tried it so far, but I can imagine at least two problems with
> > > that:
> > >
> > > 1. If I leave AFS running although I shut the network down and
> reconnect
> > > the machine the next morning, how long will it take until the AFS
> client
> > > recovers? And will it recover at all after such a "long" period of
> > > network outage?
> >
> > My laptop has a 1.0.4 client pointed at the 1.0.4 server in my house. I
> > walk off the network, then come back hours later, rejoin the network,
> says
> > fs checks, and I'm back in business.
> 
> As all of you suggest this, I will try and see how it works. I'll keep
> you posted.

Just for the record, here's my experience with that approach. Works
really well - better than I expected, to be honest. It takes around 8 to
10 minutes until my machines times out for our cell's 5 DB servers and
the varying number of fileservers it has contacted lately. After
reconnecting to the network, "fs checks" brings you back in business
immediately - just as you said. The latter also works fine if I connect
via a phone line from home in the evening and thus have a different IP
address than during the day while being connected to the LAN in the
office.

So thanks to all of you who recommended to do it that way. It really
helps me a lot! Nevertheless, with regard to the original problem, this
is still a kind of workaround. One should be able to shut down AFS
properly IMHO. From what I read in other messages on this list, this
seems to be related to the ac kernel patches. I don't know if the SuSE
kernel sources also contain these patches, although I guess that yes.
How can I quickly find out?

Apart from that, I will most likely switch to 2.4.4 soon (with SuSE
7.2). In case this makes any difference, I'll let you know.

> I've tried the following: Delete all the V* files in /usr/vice/cache
> each time before starting the AFS client. And indeed, I haven't had
> problems when shutting AFS down again since then - at least so far! (Of
> course, this really minimizes the benefit of the cache, as it only
> contains stuff that was cached since the last startup of AFS.)

Forget that one. I guess this only helps if the cache doesn't get filled
up too much during one session (i.e. between starting and stopping the
AFS client the next time). After using the the AFS client heavily the
other day, so that approx. 60% of my 100 MB cache was filled, I had the
same problems when trying to shut down AFS. Just for completeness ...

Markus

-- 
Markus Alt
IBM Lab Boeblingen, Germany
altmark@de.ibm.com