[OpenAFS] clarification on client caches

Wes Chow wes@woahnelly.net
Tue, 7 Sep 2004 17:56:59 -0400


> >>so what's different:
> >
> >3 cases...
> >
> >Debian Sarge with custom compiled vanilla 2.4.27 kernel.  MEDIUM
> >options with initial cache size of 50 gigs -- afsd segfaults
> >
> >Same Debian Sarge with MEDIUM options with initial cache size of 50
> >megs.  After boot, resize to 50 gigs -- working
> 
> you can force the files, chunksize and cachesize instead of letting it 
> calculate best fits for you. i suspect there is some number of files over 
> which you're losing.

Yet another twist... it turns out that another difference between
these setups and my previous tests is the cache size in cacheinfo.  I
originally tested with 30 gigs, which caused clients to hang on first
access to /afs.  With the above 3 setups, the cache size is 50 gigs
(for no apparent reason, I decided to try a different number).  It now
seems to work with 50 gig caches.

I've also tried 130 gigs (hangs on /afs access), and 100 gigs (works).

What does this mean:

"Be carefull, with certain cache sizes afsd crashes on startup (Linux,
Tru64Unix at least)."

http://grand.central.org/twiki/bin/view/AFSLore/AdminFAQ#3_33_Tuning_client_cache_for_hug


By the way, when I say "works", I only mean that I can list some
directories and I can cat a few files .  My tests aren't any more
extensive than that.

Wes

-- 
http://www.woahnelly.net/~wes/          OpenPGP key = 0xA5CA6644
fingerprint = FDE5 21D8 9D8B 386F 128F  DF52 3F52 D582 A5CA 6644