OneAgent on Solaris: What happens to the host when single java process dies?

[anthony_kalind1]

We are wondering what happens to single process Solaris box when the java process dies?

Can create a problem if the process dies no matter the reason?

When killed the box is reported as dynatrace not being installed which makes sense as the the agent is injected into the java process runtime.

However our client seems to think they have a scenario where a single process box might have a process die and monitoring will not alert them to any change in that case.

Could you explain or point us to some documentation on how Solaris monitoring works if a java process dies?

Extent of monitoring?

[Julius_Loman]

The host goes offline in this case. I suggest running just a dummy java process at the host just to keep the host monitoring online.

[Julius_Loman · Accepted solution]

With default settings problem will be created after the process dies or it shut down and there are request coming to this process (Apache HTTP server in this case). If you need to raise a problem if the process dies for whatever reason, you can configure it at the process group level, see here for details:

https://www.dynatrace.com/support/help/how-to-use-dynatrace/process-groups/monitoring/process-group-...

And choose "If any process becomes unavailable".

This will alert you when the process dies or is shut down. After last process on the host shuts down, you lose also host metrics for that Solaris host.

[slawek_michalik · Accepted solution]

Hi Anthony,

Per my understanding, in case of Solaris, OneAgent runs limited - meaning it's just liboneagentproc.so begin added to monitored process, which is java in your specific case - meaining no watch-dog, system-level monitors, etc. in the background

Agent gets injected to process, thanks to dynamic preload.

What does that mean for you?

Imagine an agent performs monitoring tasks as an additional thread of particular process - if process dies (no matter the reason), agent dies as well - so host stays unmonitored. Per your question regarding alerting - you'd consider @Julius L. hint, which seems to be the easiest way, or just refer to https://www.dynatrace.com/support/help/shortlink/api-events#post-events if you need more customer specific events. The second should work as simple rest call invoked by scheduled sample-availability-check.sh

Best,

Slawek