chas williams - CONTRACTOR wrote: > In message <[EMAIL PROTECTED]>,=?windows-1252?Q?Martin_MOKREJ=8A?= writes: > >> I have problem to get running a huge cache on an SMP linux box. >>The same cache size works fine on non-SMP boxes using same ext2 filesystem. >>On the problematic host, I can start using 3.6GB cache, but not 36GB large. > > > i cant explain why either would work. the current code to compute cache > fullness overflows values around 20GB which would limits this cache to > something like 16GB. to play it safe, the 18GB is the largest cache > i would try with the 1.4.0 release. this will likely be fixed in a > later release. > > what are you trying to cache?
Some files of size up to 2-3GB each. > > >>** Cache entry @ 0xde6d2d64 for 163.536870913.1.1 [phylo.natur.cuni.cz] >> locks: (none_waiting, write_locked(pid:8017 at:54)) >> 6510615555426900570 bytes DV 0 refcnt 2 > > > this value is disturbing. Would you send me a patch so that afsd would printk some debug text on it's startup? Few times I was playing with the afsd process, I managed to get afs started and mounted, but any command, like /bin/ls did not return back to the parent commandline shell. Basically, the machine had to be reset. Yes, I could login through ssh many times, bout could not efectively execute any single command. This is probably a different problem compared to the startup/mount problem, but just wanted to mention this too. I think I need printk patches if this problem should be traced. M. _______________________________________________ OpenAFS-devel mailing list [email protected] https://lists.openafs.org/mailman/listinfo/openafs-devel
