On 07/01/11 19:40, Drunkard Zhang wrote:
My configuration: cache_dir coss /mnt/c/72 10240 max-size=524288 max-stripe-waste=32768 block-size=4096 maxfullbufs=10 cache_swap_log /mnt/s/%s /mnt/c/72 is a file on btrfs + SSD. The btrfs is created by: "mkfs.btrfs /dev/sdb1 /dev/sdc1", so it will spanning across two SSDs. But squid did not write anything to disk, here's info in cache.log 2011/01/07 14:36:42| WARNING: failed to unpack meta data 2011/01/07 14:36:42| storeCossWriteMemBufDone: got failure (-6) 2011/01/07 14:36:42| FD 9, size=1048576 2011/01/07 14:36:42| WARNING: failed to unpack meta data Why? squid can't work with btrfs? or SSD? or my way using it?
COSS swaps its data to and from disk in slices of 1MB (as you can see in "FD 9, size=1048576" size= is in bytes). Check that your SSD filesystem can handle files written across the disk boundary without corruption.
Squid handles one cache_dir per spindle (in this case per SSD) much better than multiple-disk filesystems. You will get more efficient results configuring each SSD as a separate cache_dir.
Amos -- Please be using Current Stable Squid 2.7.STABLE9 or 3.1.10 Beta testers wanted for 3.2.0.4
