Lin Jui-Nan Eric
2008-Apr-17 06:06 UTC
Lots of FIN_WAIT_1 TCP Connections and deadlock in zoneli state
Hi All, I am running squid as reverse proxy on FreeBSD 7.0-R amd64. After running for a while (~ 8 hours), the throughput degrades to very very low rate. I found the squid is in "zoneli" state and is already a bug report on http://www.freebsd.org/cgi/query-pr.cgi?pr=106317 But after more investigation, I found there are lots of TCP Connections in FIN_WAIT_1 and never close. # netstat -an | grep FIN_WAIT_1 | wc -l 823 # netstat -an | grep FIN_WAIT_1 | head tcp4 0 273 192.168.1.1.80 61.62.121.195.64739 FIN_WAIT_1 tcp4 0 13448 192.168.1.1.80 163.22.44.102.58775 FIN_WAIT_1 tcp4 0 33304 192.168.1.1.80 163.32.61.253.2273 FIN_WAIT_1 tcp4 0 48416 192.168.1.1.80 24.78.35.230.61059 FIN_WAIT_1 tcp4 0 33580 192.168.1.1.80 140.130.102.8.24673 FIN_WAIT_1 tcp4 0 32009 192.168.1.1.80 163.23.57.133.38863 FIN_WAIT_1 tcp4 0 78112 192.168.1.1.80 122.146.0.210.21639 FIN_WAIT_1 tcp4 0 32046 192.168.1.1.80 140.130.102.8.23723 FIN_WAIT_1 tcp4 0 32120 192.168.1.1.80 140.128.194.130.56875 FIN_WAIT_1 tcp4 0 33580 192.168.1.1.80 163.21.235.230.33900 FIN_WAIT_1 Is there a timeout timer for the FIN_WAIT_1 state? Thank you for any suggestions!