Can tc server instrumentation replace the Spring Instrumented JARs?

[dan.baumann]

I'm evaluating tc server Spring Edition in combination with vFabric Hyperic.

According to this knowledge base article Spring Insight Operations is the recommended migration path from the Instrumented JARs. I was under the impression that tc server (or the Spring Insight Operations agent) would therefore export via JMX the same set of Spring Framework metrics as the Instrumented JARs (amongst others).

But in my evaluation setup, tc server does not expose any Spring-specific metrics via JMX (tested with JConsole).

So I'm wondering whether my expectations (tc server performs the same instrumentation at runtime that the Instrumented JARs received at build time) are valid after all.

Thanks for your input!

Dan

[Andy Wilkinson]

Insight Operations isn't an exact metric-for-metric replacement for the instrumented JARs so you shouldn't expect to see the same JMX-based metrics being exposed as you would with the old instrumented JARs.

[dan.baumann]

Thanks for the clarification, Andy.

So if I want to track the ExecutionsPerSecond of all my application's @Controller beans in Hyperic (or any of the other Spring Framework metrics), I need to include the old instrumented JARs in my application. That's fine until the old instrumented JARs will hit end of life in about 18 months.

I always thought of the Spring Framework metrics as a great advantage of using Hyperic with Spring. But as nice as Insight Operations really is, I don't really see it as a replacement: In my understanding, Hyperic is for long-term monitoring and tracking of metrics plus alerting (targeted at operations types), while Insight Operations is great for understanding single incidents or spikes (targeted at developer types). I didn't expect either of those technologies to lose any features.

What added to the confusion is the fact that both technologies use instrumentation (instrumented JARs at compile time, Insight Operations Agent at runtime), but, as it turns out, for different purposes. I was therefore under the false impression that Insight Operations does at runtime what the instrumented JARs do at compile time (in addition to the instrumentation required for Insight Operations), and that Insight Operations could therefore be a replacement for the instrumented JARs.

The fact that Insight Operations is supposed to be the recommended migration path from the Instrumented JARs didn't help to clear this up either (our VMware representative also seemed to be struggling with this). The (otherwise excellent) documentation really should address this.

To be honest, I still don't know how Insight Operations is supposed to replace the instrumented JARs (or "be the recommended migration path"). But at least now I know what's possible and what isn't.

[Andy Wilkinson]

Thanks for the feedback. I've passed it onto the Insight team and a few others so that they're aware of the problem.

You can integrate Insight Operations with Hyperic which, among other things, enables collection of metrics from Insight giving you the long-term tracking and alerting that you're used to.