[OpenAFS] Possible bug in OpenAFS 1.4.1-2
Mon, 04 Sep 2006 21:17:17 +0200
Content-Type: text/plain; charset=UTF-8
I would like to report a rather nasty change in behaviour between 1.4.0
and 1.4.1-2 versions of OpenAFS, as packaged by Ubuntu. The problem
manifests itself as repeated attempt to connect to an unreachable cell
EVEN IF all operations within /afs are done within another cell's
I have a laptop that runs afs client with two cells configured: first,
set as home cell, is rikanise.net, my private cell; it uses DNS
information only. Second cell, used at my workplace, is statically
configured in CellServDB file, since we didn't (yet) get around to put
the info into DNS.
At work, both cells are reachable, even if connection to my home cell
frequently drops, due to NAT problems. At home, I cannot use my
workplace's cell, as it is firewalled and currently unavailable outside.
Up until the upgrade, I could work with content within my home cell
without problem even if the other cell - mounted automatically by the
daemon - was unavailable. running dmesg would show several lines of
'connection to volume location server is broken', but any timeouts
associated with those were masked.
After the upgrade, the behaviour changed dramatically: _any_ operation
within /afs structure would hang for (I assume, I didn't actually
measure that, but still, it was bloody long time...) 2x180 seconds, if
the connection manager would have problems contacting either cell!
In short, any operation within my home cell, done _at_home_ was
predecessed with very long timeout during which cache manager was trying
to contact the _work_ cell.
Removing information about my work cell from CellServDB and restarting
the client removed the timeout, so I'm pretty certain that I can rule
out misconfiguration at my home cell server's side.=20
Some basic info: Ubuntu Dapper, i386, OpenAFS ver. 1.4.1-2 from stock
package; I will check version directly downloaded from official site
Unix Systems Administrator,
Wy=C5=BCsza Szko=C5=82a Zarz=C4=85dzania i Bankowo=C5=9Bci w Krakowie.
Content-Type: application/pgp-signature; name=signature.asc
Content-Description: To jest =?UTF-8?Q?cz=C4=99=C5=9B=C4=87?= listu
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v188.8.131.52 (GNU/Linux)
-----END PGP SIGNATURE-----