Hi Haytham, On Fri, 18 Jan 2008, Haytham KHOUJA wrote:
Hello, I don't advise you to use partitions as separate cache_dirs. You will not get performance enhancement since you're still working on the same physical disk and same SCSI controller.
Its not 2 partitions on 1 physical disk, I have 2 scsi disks... controller however is same.
Moreover, 1GB for cache_mem is way to much if you only have 3GB of total RAMs if you have 8x8GB cache_dir. Lower that to around 256 MB ~ 512 MB and try.
Let me try that... I still have free mem though. by the rule of thumb, we need 10MB Ram for indexing per GB of cache. that sums up to 8 x 8GB x 10MB/GB = ~650MB RAM for cache index + 1GB for Cache_mem = ~1.7GB. so I'd still have ~1.3GB of RAM for system operation.. I'll lower the cache_mem, however, to try..
Thanks. Manoj
Hi all. I'm having an issue with our cache. its a 2.6S17 on centOS 4 x86_64 with 3GB Ram 2x36GB scsi disk for cache_dir and separate SATA 80GB for system. It seem to flush the mem cache for no apparent reason.. nothing in the cache log around the time of occurance... please see the graph in the following link.. it has done it twice in 10 days. we've not encountered such issue. relative squid.conf settings follows: cache_mem 1024 MB maximum_object_size_in_memory 16 KB cache_replacement_policy heap LRU cache_dir aufs /sda1/cache0 8000 20 256 cache_dir aufs /sda1/cache1 8000 20 256 cache_dir aufs /sda1/cache2 8000 20 256 cache_dir aufs /sda1/cache3 8000 20 256 cache_dir aufs /sdb1/cache0 8000 20 256 cache_dir aufs /sdb1/cache1 8000 20 256 cache_dir aufs /sdb1/cache2 8000 20 256 cache_dir aufs /sdb1/cache3 8000 20 256 maximum_object_size 18432 KB cache_swap_low 95 cache_swap_high 97 access_log /usr/local/squid/var/logs/access.log squid cache_store_log none logfile_rotate 7 2 36GB scsi disks are mounted on /sda1 and /sdb1 as /dev/sda1 on /sda1 type reiserfs (rw,noatime,notail) /dev/sdb1 on /sdb1 type reiserfs (rw,noatime,notail) I can see lots of entries in cache.log that looks like: 2008/01/18 09:24:18| storeLocateVary: Not our vary marker object, 1B384B5F6E91A0FE402D4DA6F9B067B0 = 'http://x.tagstat.com/js/announcements.js', 'accept-encoding="gzip,%20deflate"'/'gzip, deflate' Any suggestions please.. Thanks Manoj --
--
