NeophyteNeophyteI also have a datasource that seeks to help automate process discovery and monitoring sitting in Security Review (for about 10 months now -- Locator code: 9YLFYE; I doubt anyone can actually see this.)
My approach was more or less reverse engineered from Zenoss’s process monitoring, where the user provides a list comma-delimited regular expressions, and the WILDVALUE reflects the regular expression as the instance. All data is rolled up into ‘# of matching processes,’ and their CPU/Mem values are rolled up as well. This was primarily important for Java process monitoring, as all Java processes will commonly have `java` as their hrSWRunPath, despite being many discrete applications sharing this hrSWRunPath.
Of course, my solution was a means-to-an-end to achieve parity with our incumbent platform at the time, and I actively discourage any of our users to make further use of process monitoring. A better pattern is figuring out how to assess the operational health of a process through querying it directly instead of predicating it upon whether or not we see it in the process list. SNMP is also suboptimal for this, as it can show a process running even when it’s zombied or orphaned. But, alas, sometimes SNMP is all we have :)
