Question

We had problems with our Oracle 10g DB which reached the CPU Usage limit. (It’s a Java Web application using Hibernate as Persistance Framework, JDBC and a c3p0 Connection Pooling)

In our monitoring we can see that there is a strong correlation between the cpu usage and the active sessions.
There is a single peak to around 75 active session (the active seesion are usually around 10) wich is probably responsible for the trouble we run into.

My question is, how can I find out what was responsible for this peak in active sessions?
Is it possible that a crawler could cause this peak?

[Edit]
We analyzed access log and found out that there is no traffic peak during that speficic time-intervall, in which the active session peaks occured.
I suppose it is a consequence of the high CPU, wich leaded to extremly long DB query durations. But it’s only an assumption.
Maybe someone has any hints or further explanation?

Thanks in advance for any hints and advice.
Ronny

Answer 1

You can verify whether a crawler caused the peak by analyzing the web server log:

• Can you see an unusally high number of requests during the peak time?
• Does the agent identification include words like bot, spider or crawler?