Visualising ConfigMgr Site Servicing from a unique perspective for fun

My DA was broken last week, so while away at the MVP 2017 SCCM Hackathon Summit, I couldn’t nudge my lab server to upgrade to the latest Technical Preview, Build 1711 shortly after Jörgen Nilsson and Kim Oppalfens clicked the button to release it to the public (shot below!), but last night I set it off, and put told one of my minion tools to keep an eye on what the Inboxes were doing during the upgrade just for the fun of it.

That’s TP Build 1711 flying out the door with a click on the tablet in front of them.

This post isn’t much more than me sharing some pictures I thought were fun to see, the tool being mentioned is InboxWatchdog which is used internally at SMSMarshall Ltd to analyse the load a Site server is under, so that we can determine how fast its queue clear-rates are (good indicators of quality of architectural design and overall health). It is worth saying that SCCM is specifically designed to handle data surges from the client estate that produce backlogs in the inboxes, which the product dutifully munches down on and mostly hives off into SQL, and that SCCM is put under serious load and left to clear everything down within 24 hours as an indication of ship-worthiness by the product team.

Not much happened in the inboxes on this Site being serviced as you’d expect, since the Site is pretty much down until the process of servicing has almost completed, with a little surge at the tail end once the Site wakes up with a new jacket and hat on, as a result the tool didn’t record many data point samples:

image

But we get to see some State Message samples showing up (Inboxes\statmgr.box\statmsgs) which produced a nice blip on the graph:

image

Here’s State Message on its own:

image

This next shot I put in for no other reason than it looks pretty rad. It shows how long the tool took to sample data while it was running. I was hitting a Server OS bug in 2015 I believe, which made the sampling rate grow exponentially resulting in the two ends of this graph diverging away from each other into a cone shape, but with Server 2016 I’m not seeing that, its pretty consistent:

image

And here’s the CPU and Memory usage, but it includes a lot of data post-upgrade, that was sampled before I stopped the tool in the morning:

image

Was a bit of fun to see what a Sites profile looked like when its servicing itself.