I am running an OpenAFS 1.6.11 client against a set of 1.6.20 servers. I am trying to find out if my client's cache is underutilized or thrashing because if it running out of entries. I've read the documentation here ( http://docs.openafs.org/Reference/5/afs_cache.html ) and the afsd page here ( http://docs.openafs.org/Reference/8/afsd.html ) but I think I still don't have the whole picture. So I'd like some help understanding the disk based cache configuration.
My cache is configured via /etc/openafs/cacheinfo with 1000000 blocks (1GB) of maximum space. afsd is started with -dynroot -fakestat -afsdb flags.
Now, what I read from the afsdb manual my chunksize by default should be between, 256KB and 1MB. I can tell by looking at the Vn files in the cache directory that chunksize is 1M, since there is no file greater than this size while a few are more than 1023KB.
I also read that my Vn files number should by default be chosen as the greater of (100, 1000000/10240, or 1.5*1000000/1024). The latter being the greater I'd expect to find no more than 1464 Vn files in my cache directory. Contrary to this, du --inodes says there are 31270 files in my cache directory. Indeed, the highest numbered Vn file is V31249 and the rest of used inodes can be attributed to parent directory entries.
So what am I missing here? How does the cache manager calculate the limit to Vn files?
Thanks in advance, -Kostas.
Description: S/MIME Cryptographic Signature