Sstrhsrtj

I am running a laravel application on a Ubuntu 14.04 digital ocean vps and I am using New Relic to monitor the server.

I got an email alert that my CPU usage was above 80%. I logged in to New Relic and now it's showing my CPU usage at 99% for 18 hours now. But when I log into my shell and run 'top' the CPU usage of the processes don't even sum up to 10%.

What could be wrong? Which are other commands I could run to check the real usage and what is using it so much? (Perhaps an infinite loop on the application?)

This is my htop result:

And this is htop after shift+K

Any links or help would be greatly appreciated.

I had a similar problem and I did run TOP command in background and redirect it to a txt file

top > top.txt &

Thank you can see which process waste your CPU

There are a few options. My preference is the program htop which isn't generally installed by default but is available for almost all distros. It gives you the same info as top but with a MUCH richer feature set. It's great for interactively finding the problem process but it's not great for scripting. If you want to incorporate something into a monitoring script the command ps -eo pid,pcpu,comm will be your friend. It prints all of the current processes along with their PID and CPU usage in a fairly easy-to-parse format.

I'm not sure what's going on, but my guess is that the individual process's %CPU column isn't telling you what you think it tells you. From top's manpage:

  k: %CPU  --  CPU usage

  The task’s share of the elapsed CPU time since the last screen update,

  expressed as a percentage of total CPU time. 

Let's say you're refreshing every 4 seconds. The total CPU time it consumed in those 2 seconds is 6% over those 4 seconds. But let's say in one of those seconds, it spurted and used 24% of CPU time. I'm not saying this is happening, but it could be.

I am happy with top here. I can see CPU usage broken down by CPU and further by user, sys, nice, idle, wait (on I/O), servicing hardware and software interrupts (hi, si). You can probably get this in htop as well. As long as you have 1 CPU that is at least 20% idle, there's little to actually worry about, unless your fan is bothering you. But if you're worried, maybe what's happening is %wait and %hi/%si are very high, in which case no individual process is having a high CPU usage, but rather, the kernel is very very busy for some reason.