My MVP Profile

Saturday, April 07, 2007

Unknown State for Ops Mgr 2007 Agent Health Service Explained

You may have seen the agent health service status of one or more of your agents go into a grey(unknown) in the Operations Console of Operations Manager 2007, like so:

That being said, the real question is 'what is the underlying program logic that causes state to unfold in this way?' I posed a question to Jakub Oleksy, developer on the Ops Mgr 2007 product team:

“What is the logic moving status for an uncommunicative agent to the grey(unknown) status in representing object state, rather than simply assuming a down state and moving state to red?”

…and he provided some very useful information on the logic behind these state changes.

His answer:

“We actually do go red and do generate an alert, just not for the computer directly.

We have a notion of a health service watcher that is monitoring the
availability of the health services. If a health service fails to heartbeat for 3 consecutive intervals, that health service and anything monitored by it (including the computer) goes grey (i.e. unavailable)

We then try to ping the computer:

If we CAN ping it, then the health service watcher stays red and we generate a Health Service unavailable alert.

If we CANNOT ping it, the health service watcher goes yellow but a Computer Not Reachable alert is generated.

We also have a state view for this:
Monitoring\Operations Manager\Agent\Agent Health State”


I do have more questions after observing this behavior again with a fresh perspective of the underlying program logic, but certainly this is a great start.


Thanks Jakub!

Comments on "Unknown State for Ops Mgr 2007 Agent Health Service Explained"

 

marius [MSFT] said ... (5:14 PM) : 

It should be added that each instance managed by health service which is marked unavailable will become unavailable as well. If this instance is then contributing to some dependency monitor, state of its contribution is defined by %gt;MemberUnAvailable< on this particular dependency monitor. Instance will not contribute if %gt;MemberUnAvailable< was not defined.

 

post a comment links to this post