A more general metric that is useful to watch for is pressure stall information for CPU, IO and Memory.

https://docs.kernel.org/accounting/psi.html

I made a Prometheus exporter for it:

https://github.com/arianvp/cgroup-exporter