When the network clients are in close proximity to the
Caching Proxy server (same building) or Caching Proxy frequently
experiences heavy stress, decrease the tcp_time_wait_interval
kernel parameter value.
If the tcp_time_wait_interval value is too high, the proxy process might
use excessive memory because too many connections are in the TIME_WAIT
state.
To test if the proxy needs to be tuned, issue:
netstat -na | nawk '{print $NF}'| sort |
uniq -c | sort -n
If the number of connections in TIME_WAIT state exceeds 1000, adjust
tcp_time_wait_interval by issuing:
ndd -set /dev/tcp tcp_time_wait_interval
5000
To view the current setting, issue:
ndd -get /dev/tcp
tcp_time_wait_interval
The default value of tcp_time_wait_interval is 240000 ms on Solaris 5.8
and 60000 ms on Solaris 5.9. Set it to 5000 ms initially, and adjust it
accordingly to match the needs of your network.
For more tuning information, refer to the Caching Proxy Administration
Guide. |