It'll choose the cache dir with the least pending disk ops, if you haven't configured any other restriction (min/max object size.)
Look at the "store directory stats" page in the cache manager and see what it says. Adrian On Mon, Aug 20, 2007, Nicole wrote: > > Hello > I am testing FreeBSD 7.0-CURRENT and zfs. > > I am using: > > cache_dir aufs /cache0 24000 32 128 > cache_dir aufs /squidcache1 90000 64 256 > cache_dir aufs /squidcache2 90000 64 256 > cache_dir aufs /squidcache3 90000 64 256 > > Cache0= standard ufs/ffs filesystem slice of da0 > /squidcache1/2/3 = a whole disk 147G (da2/da2/da3) utilizing zfs. (basicly > in JBOD) > > (it's new, it's cool, but how does it work with squid...) > > Normally, (when all using ufs/ffs) all the squidcache1/2/3 dirs start filling > first and the cache0 dir fills slowest. However, with the 3 using zfs, cache0 > is filling much faster. Is this indicitive of the 3 cache_dirs using zfs > are somehow slower, so its using cache0 more? >
