[OpenAFS-devel] Large-Cache Initialization (Proposal)

[Sam Hartman]

>>>>> "Todd" == Todd M Lewis <utoddl@email.unc.edu> writes:

    Todd> Derek Atkins wrote:
    >>  So I tried (and failed) to initialize a 3GB AFS Cache last
    >> night.  I set my Linux (RH 6.2/Linux 2.2/ext2-fs) box to the
    >> task around 8pm or so last night, and at 9am this morning it
    >> was still chugging along.  I gave up, newfs'd the cache
    >> partition, reset the cache size to 325MB, and the
    >> initialization was done in a few minutes.
    >> 
    >> I'd like to fix this.
    >> 
    >> My guess (I haven't done any tests yet, but I will) is that the
    >> directory is getting so large that the cache-file creation is
    >> taking more and more time as the cache gets bigger.  My plan
    >> would be to break up the cache directory into multiple
    >> sub-directories, to try to keep each directory to a
    >> 'reasonable' size (say, ~2000-2500 files?)
    >> 
    >> This would definitely require changes to afsd to build/scan a
    >> directory hierarchy instead of a single directory.  I think it
    >> might also require changes to 'fs' (for setcachesize).
    >> 
    >> I wanted to get feedback about my proposed approach before I
    >> spend the time to actually build such a system.

    Todd> It seems a shame to mess up otherwise good clean working
    Todd> code to compensate for a file system that isn't up to the
    Todd> task. Seems like switching to ReiserFS, JFS or something
    Todd> along those lines (assuming that that would work) would be
    Todd> better -- in this case...

Which don't exist for Solaris or BSD.