why there are 20000 connections on same port?

mvairavan · 01-17-2013, 12:36 PM

Hi,

In the head node of our cluster there are almost 20000 connections open on the port number 7625.
And all these are in TIME_WAIT state.

And on the port http, there are almost 10000 connections open and even all these are in TIME_WAIT state.
Why is this happening?

And when i try to ping a web address i am getting

Code:

ping: sendmsg: Operation not permitted

Is this because of the above problem ? How do i solve this?

There is no packet loss or problem if i ping from other nodes on the cluster to any other nodes apart from the headnode. So clearly the problem is with the headnode. I tried restarting it still the same problem.

How do i solve this.

Thanks in advance

gilead · 01-24-2013, 06:11 PM

If you mean what is the TIME_WAIT state, there is a brief description here. If you want to know what processes are listening on those ports you can do this:

Code:

fuser -n tcp 7625
fuser -n tcp 80
netstat -tunape | egrep '7625|80'

The ping not being permitted could be due to a firewall or sysctl setting - are you familiar with checking those? For example:

Code:

sysctl net.ipv4.icmp_echo_ignore_all