This content has been marked as final. Show 1 reply
Problem concerning TCP has to do wait TCP connections going into TIME_WAIT state.
To see if there are any sockets in the TIME_WAIT state you can use
When the number of TCP connections in TIME_WAIT state approach the maximum number of file descriptions your application throughput will suffer degradation
netstat -a | grep TIME_WAIT | wc -l
To determine the time wait interval you can use:
Try setting this value to 60000 milliseconds
/usr/sbin/ndd /dev/tcp tcp_time_wait_interval
Note that some values cannot be set using the ndd command but must be changed in /etc/system.
/usr/sbin/ndd -set /dev/tcp tcp_time_wait_interval 60000
To check which commands are read-write or read-only you can use
On solaris 10 you might also want to change ipc1_conn_hash_size (controls the hash table to locate the
ndd /dev/tcp \?
TCP socket data structure in the kernel) by changing the /etc/system file
Note that the parameter tcp_conn_req_max_q controls the maximum allowed number of completed connections
waiting to return from an accept call. This parameter should only be increased if the system is dropping connections.
To determine the number of connection drops you can use:
If the system is dropping connection try increasing the value to 1024.
netstat -s | grep tcpListenDrop
If your machine is to host a HTTP server, it is wise to increase the open file descriptor hard and soft limits, for example to 4096
More information on tunable parameters can be found here: http://download.oracle.com/docs/cd/E19253-01/817-0404/
set rlim_fd_cur=4096 set rlim_fd_max=4096