Simon Wilkinson wrote:
There are a number of issues that you will encounter.
Firstly, as you note, there is a danger of the root disk filling. The reserved
percentage won't help with this - a user process could fill 95% of the disk,
and the remaining 5% still not be large enough for the AFS cache. Sadly, the
Unix cache manager is not particularly robust when it is unable to write to its
cache - the machine will likely panic if this occurs.
Secondly, there are almost certainly going to be performance issues in sharing the cache
with other data. ext3 is not particularly performant compared to ext2, and sharing the
ext3 journal between your cache filesystem and the rest of the system won't aid in this.
The fact that we goofed the setting of the "don't update the access time" flag
on cache accesses won't aid you either.
That said, I know of lots of sites that do run with their AFS cache on the
shared root, so you'll probably be okay. Things just won't be quite as robust,
or quite as speedy, as they might be.
OK, thanks for the info. That reassures me. I may tweak the reserve
percentage or use a loop-back mounted FS as Andrew suggested. The
machines in question have 1TB drives, so 5-10% is quite a bit of space.
Thanks,
Jason
_______________________________________________
OpenAFS-info mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-info
