hi,
I’m currently running with an issue whereby we are experiencing very poor access speeds to our Informix database. Connections or requests to the DB are taking in excess of 2/3/4 minutes during peek periods during the day. This has only just started to happen but so far we have been unable to find a root cause or identify anything that has changed.

What we have done is to open up another TCP port on the server and when we send a request to this TCP port then we get a response from the DB within usually 2/3 seconds which is acceptable. We have created a couple of scripts that perform a simple query through each TCP port and we can see that it’s something going on that delays the request the first TCP port.

Now although the system is fairly heavily utilized during this period because we can get an acceptable service through a different TCP port we know the issue is not disk/memory/cpu etc. It is not a network problem external to the server however has anybody experienced anything like this before?

One option I am currently examining is a config called somaxconn which is currently set to 1024 on the server. I’ve done some digging and it appears that it will queue up to this value of TCP requests however I do not know if this would contribute to the issue? (I have been running a netstat looking at established connections during the period and the max they go up to is just over 1100 – however because we have a 2nd tcp port open then does this let me go over this connection limit? (or even is it a limit?) or could this issue somehow be a limit between unix and Informix on a tcp port?

Now at no stage have our test queries ever timed out and we’ve even seen them take up to 5/6 minutes to return the result – however I am now in the position that I cannot remove the 2nd TCP port as I have a few business critical queries (simple) that get fired through that port that give a response in the required time and can let the business trade.

Any ideas?
Thanks
ian