Finding Cause of dynaTrace Server Resource Usage Spikes

So - MPS - is not related with the number of Measures you have in your System Profile. We know that this name is confusing - there is some historical background on this. It means how many data packages we receive from our agents. These packages contain data form PurePaths or from non-PurePath measures.

If you see a Spike in these MPS it means that more data is captured. I would assume this is due to a higher peak in traffic. If you have multiple Collectors you could look at the Collector Health DAshboard and see from which collector this traffic comes in. If you then know which agents are connected to that collector you can look at the PurePaths that came in in that timeframe from these agents. If you have that information you can check whether you have a larger number of PurePaths than average coming in or whether the size of PurePaths is just so much bigger. Maybe there are some batch jobs running during off-business hours that generate very long PurePaths? That is a very common scenario. So - check it out and let mek now what you fin

Thanks for the response. A couple of things.

I definitely did check out the Deployment Dashboards page beforehand. But, it appears to contradict other posts in the forum. Is MPS then the actual measures from System Profile > Measures being shown per second? Or are they some other measurement?

It's definitely not the PW cleanup. This occurs randomly during all hours.

We're still on 5.6 ( got to convince mulitple business units to let us uplift to 6.1! ), so everything is being handled by the pre-split dT Server. We have intructed all users to use very short timeframes unless absolutely necessary. Finally, many of these spikes occur during non business hours when there are essentially no dT users, so it's not the dashlets.

The Realtime Analyzer Queue Size actually seems stable. But, I do see a spike in Server Traffic, which does make me think it is indeed more load coming in. So there lies my question. Is there an easy way to see where this load is coming from?

Hi

for details on MPS and other indicators on these health dashboards check out Deployment Dashboards MPS in general indicates how much data is collected from your individual agents. If that number spikes you know you have more traffic on your agents - or - they are hitting some code in your application that causes dynatrace to capture a lot more data, e.G: excessive number of Exceptions, DB, ...

As to your periodic spikes. Do you have some periodic tasks running? There is one task that by default executes at 2AM in the morning. it is the performance warehouse cleanup task which is taking care of performance warehouse data aging. If that is the spike you see then I can tell you that that is expected. if you see other spikes than it be related to e.g: spikes in application traffic in combination with some heavy-weight Business Transaction analysis maybe? Check out the real time analyzer queue. If that is also spiking you know that the CPU and Memory spike is related to PurePath analysis. Otherwise it is more likely related to these background tasks. Another option could be that you have end users refreshing their dashboards very frequently. Most dashboard requests are now handled by the frontend and no longer the backened dynatrace servers. But some requests, e.g: Visit-based Dashlets or Incidents are still handled by the backend dynatrace server. So - if you have end users that have their dynatrace clients open and refreshing these dashlets using a very long timeframe it would also impact CPU and Memory

Andi