OS: Windows Server 2008, SP2 (running on EC2 Amazon).
Running web app using Apache httpd & tomcat server 6.02 and Web server has keep-alive settings.
There are around 69,250 (http port 80) + 15000 (other than port 80) TCP connections in TIME_WAIT state (used netstat & tcpview). These connections don't seem to close even after stopping web server (waited 24 hours)
Performance monitor counters:
- TCPv4 Active Connections: 145K
- TCPv4 Passive Connections: 475K
- TCPv4 Failure Connections: 16K
- TCPv4 Connections Reset: 23K
HKEY_LOCAL_MACHINE\System \CurrentControlSet\Services\Tcpip\Parameters does not have TcpTimedWaitDelay key, so value should be the default (2*MSL, 4 mins)
Even if there are thousands of connection requests are coming at the same time, why windows OS is not able to clean them eventually?
What could be the reasons behind this situation?
Is there any way to forcefully close all these TIME_WAIT connections without restarting windows OS?
After few days we app stops taking any new connections.