Hi, I run into issue with socket in the server where I use squid. Below is my server configuration: Squid 2.6. stable 4 - 20061009 Linux AS server version 4, update 3 Dell 2950, 16 GB RAM, 2 dual core CPU, six 164 GB SAS disk, RAID controller. RAID 0 implementation with prefetch read, write back policy. The squid is configured in reverse proxy mode with 120 domains, 8 origin servers. (origin server is apache 2.2.0 on linux) Request object size = 1064 Bytes Number of unique object = 1 million Object's TTL=5 minutes Persistent connection to server = on Persistent connection to client = off The squid has 32768 max file descriptor 90+% hit ratio. I send a load of 4000-5000 request/second to the server. After about 5 hours, the ability to serve 5000 requests drops and the cache.log is filled with error 'commBind: Cannot bind socket FD 204 to *:0: (98) Address already in use I suspect the server run out of socket, so I run netstat to capture the open/pending to close socket and there I see one origin has approximate 8000 connection in TIME_WAIT state .. looks like the socket is close but the OS keep it open for a bit longer in case the data is not transmitted complete. That seems to be one problem so I want to reduce the TIME_WAIT. I did research on how to reduce the TIME_WAIT and one recommendation is to change the tcp_fin_timeout to 30 (original value is 60), tcp_sack=0 (original value=1), tcp_window_scaling=0 (original value=1), tcp_keepalive_time=2400 (my value=1800). Any recommendation or suggestion are greatly appreciated ? Or my issue is entirely different? since the error in cache.log says it tries to bind to port 0. Seems pretty odd, but at least the excessive number of TIME_WAIT connection on the server bothers me. Thanks, Khanh
