Hi,
I've been playing with a CLI, JMX and a few other things. At work I am manually extracting and monitoring various bits of JMX data in order to diagnose problems with Tomcat configs and applications. There's lot of information available in the JVM & Tomcat which could be polled periodically, (theoretically comparatively cheaply), in order to determine if there is a problem developing or occurring. Some of this information could be collated and presented in a structured format to enable more easy analysis. Tomcat could then self-diagnose some common problems. Rainer & I had a lengthy conversation at the Apache Retreat last weekend and extrapolated, coming up with more ideas: - Periodic snaphots of key indicators, e.g. Connector backlog, error counts, resource pool usage. - A 'black box recorder' mode which can be enabled to log key data - Periodic inspection of threads to warn about blocking, deadlocks - A simple web UI (like the /manager) which presents & collates this info and does some simple analysis It should also be possible to make some educated guesses about the sources of common problems by doing statistical analysis of thread stacks. There are products on the market and app in the JDK which instrument apps and perform profiling, but we concluded we can simulate some of this with little performance cost. There's a huge gap between understanding basic info, and understanding what the JDK tools output. It should be possible for Tomcat provide an intermediate level and help users improve their own analysis. No silver bullets, just a step up from the basics. Thoughts? p (@Rainer did I forget anything?)
0x62590808.asc
Description: application/pgp-keys
signature.asc
Description: OpenPGP digital signature
