We've been conducting some performance testing on our RStudio Connect server and have found a weird behavior on the Admin - CPU Graph.
It appears the values are wrong. Specifically they are out by a factor that relates to the number of cores on the machine.
For example, when we run a (long) intensive task on a single core on our 8 core server I would expect the CPU usage to be at 12.5% (i.e. 100% / 8). However it sits at 1.57% which is conveniently 12.5% / 8. We then tested this with concurrent sessions running and see similar results. So with 2 processes run, CPU caps out at ~3.1%, 3 processes, ~4.7%.
Thanks for letting us know what you're seeing. The best recommendation I can give you for now is to file a support ticket so we can keep you updated. Out of the diagnostic information on that list, the host information will likely be the most important thing to include.
We have confirmed that there is an issue with the CPU graph in the 1.7.2-1.7.6 releases. As you suspected, we were incorrectly dividing the utilization number by the number of CPUs (cores).
There is no workaround, but the problem will be resolved in the next release of RStudio Connect (likely called 1.7.8).
