CPU util vs. oracle connection get refused

[Hanry Zhou]

Oralce users got the problem of connecting Oracle, and got the message of "connection refused". They suspect this has something to do with CPU utilization when the error came up.

I don't feel that was the cause of the problem, but what I should tell them, and anybody has the same experience?

[Jakes Louw]

I think you should try and match any message in the oracle alert log to any event in the system log, as a first step. As a second step, I would start collecting CPU stats using SAR, or PerfView if you are licensed for it. Then if and when they complain, you'll be able to see what was happening at that time.
Otherwise, anything you say is likely to be guesswork....

[rick jones]

Connection refused could be any number of things, I think you should start a tcpdump trace somewhere looking for TCP RST segments from the server. If you capture the whole segment, there will, in ASCII, be the reason for the RST segment which will help you track-down the reason.

[Hanry Zhou]

Whenever the user has any type of issue, they always have suspcious on what is happening on system, even though most like are application problems.

So, in this case, I don't think the cpu utilization will be the cause of the error messages. I just wanted to find out from you what i should response to them, and if any of you have the experience that CPU util will effect to connect to oracle? or what are the possible causes of the problems?

[rick jones]

Connection refused could be that there was no listen endpoint open at the time. It could be that ipfilter had rules that rejected it. Or an intervening firewall.

[Jakes Louw]

Hanry, in my experience, I have never seen that. More likely an I/O issue, and I suggest you start learning some basic Oracle: how to connect from a remote machine (sqlnet) and how to connect locally (sqlplus). Also, have a look at comparing TNSPING to normal ping results. If normal ping reports slow speed ping replies, you have a network/card congestion issue.
If that looks good, but TNSPING is slow, you have a DB authentication issue. If TNSPING and normal ping are OK, but a simple SQLPLUS connection is very slow or times out, you could have I/O congestion issues. Only then will it be time to check disk I/O stats.
Like you said: It's always easy to blame the OS!!!