Clients consuming all TCP connections
Summary
We seem to be having an issue where tcp connections keep getting maxed out and then no further queries via tcp are possible. We have usually identified a few clients that have lots of connections in CLOSE_WAIT state and then firewall them off.
BIND version used
named -v
BIND 9.16.25-RedHat-9.16.25-1.el7 (Extended Support Version) id:3e14423
Steps to reproduce
start bind
What is the current bug behavior?
client consumes all tcp connections and bind no longer responds. udp still works fine
What is the expected correct behavior?
bind should close the sockets properly?
Relevant configuration files
# grep tcp /etc/named/named.conf
tcp-clients 2048;
Relevant logs and/or screenshots
# netstat -padn | grep CLOSE_WAIT | grep 123.456.789.123 |wc -l
274
Will attach a tcp dump if needed.