Focus on CPU

Hello there. We're going to take a quick look at how we would go about investigating a high CPU load situation within Foglight. As we can see here, there's a couple of areas that we might want to investigate this. The first one I'm going to show you is around an alarm showing me that the CPU on the VM here is suddenly spiking. It's not within the baseline we'd normally expect to see for this time of day, and therefore its traffic light is red there that we need to go and investigate that, but at this stage that could be related to some high amount of workload on the instance, or it could be related to something running on the guest operating system, or indeed it could be something to do with like the limits at the hypervisor level that we need to look into. But we'll just investigate a couple of different areas that we can view this. So first of all, let's just check the workload. We can see when we hovered over the workload that the workload suddenly increased, so that's the first sort of indication that something's not quite right. And as we drill into the performance investigative dashboard here, we could have got there directly by clicking on SQL PI, we'll see that we over the last few minutes have had an increase in workload. Let's just click on SQL Statements View just to have a look at what sorts of queries are running in that particular area, and indeed we could drill down into a particular database if that was the problem in question. So we can see the workload and we can see that we have-- for whatever's running, we've got an amount of other wait some memory wait. CPU usage is increased but also CPU wait, and we can see the top queries at the bottom, but to focus on CPI in particular, we can click on the CPU Resource button at the top and that will resort these statements into the queries that are using the most CPU. So these are my top queries sorted by the ones with using the most CPU usage, and we could click into any of these particular queries, and we might want to tune the query or analyze the execution plan for that particular query. We could just have a little quick look at the execution plan, and this is where we could identify whether the high CPU was related to a particular operator on the execution plan. So that's a quick way of just seeing what the top queries are. As I said, it may not be the queries that's the problem, it may be the operating system. So the other step we could take if we went back to the database's dashboard is we could drill into the actual host here itself. I've just got the view here where we can see that over the last few minutes the CPU started to spike. We've got a health history issue with high CPU, and indeed we could then drill into the CPU processes on the guest level just to see how long this problem would have been taking place and which particular processes are taking up all the CPU. And we can see that SQL Server.exe is the top process, but there are some other processes there using a lot of CPU, and the change rate over the hour, so 50% increase in workload there. So that's just a quick overview of drilling into a CPU-related issue. Thanks for listening.