Re: [squid-users]Memory leak problem on epoll i/o squid on IA64

[David Nicklay]

Hi,

MUTHUKUMAR KANDASAMY wrote:
Hello all ,
In the compilation of squid ,I used  cache_mem 1200 MB cache_dir null 
The cache line looks fine, I doubt that is causing you problems

fs.file-max = 16384
That should probably be higher, like 32786 or something higher...

net.ipv4.ipfrag_low_thresh = 900000
net.ipv4.ipfrag_high_thresh = 1000000
net.ipv4.ipfrag_time = 45
net.ipv4.tcp_rmem = 2000000  2250000 2500000
net.ipv4.tcp_wmem = 1000000  1250000 1500000
net.ipv4.neigh.default.gc_thresh1 = 1024
net.ipv4.neigh.default.gc_thresh2 = 4096
net.ipv4.neigh.default.gc_thresh3 = 8192
net.core.rmem_max = 1500000
net.core.rmem_default = 1500000
net.core.wmem_max = 1000000
net.core.wmem_default = 1000000
What the heck?  Why are your settings set to such insane levels?  The 
first number in net.ipv4.tcp_wmem and net.ipv4.tcp_rmem would cause you 
problems.

tcp_wmem - vector of 3 INTEGERs: min, default, max
        min: Amount of memory reserved for send buffers for TCP socket.
        Each TCP socket has rights to use it due to fact of its birth.
        Default: 4K
tcp_rmem - vector of 3 INTEGERs: min, default, max
        min: Minimal size of receive buffer used by TCP sockets.
        It is guaranteed to each TCP socket, even under moderate memory
        pressure.
        Default: 8K
Taking a wild stab, and saying 256 file descriptors open, that means 
something on the order of:

256 * (2000000 + 1000000) = 768 MB (and that is MINIMUM even under 
moderate memory pressure)

Lower all of your numbers down to a sane level....

--
David Nicklay
Location: CNN Center - SE0811A
Office: 404-827-2698    Cell: 404-545-6218