We were directed to this bug by our Portworx support team, and believe we're seeing the same thing.
We are running Ubuntu 16.04 with the hardware enablement kernel. Most of our nodes are running 4.15.0-140, and like Peter notes above, the I/O performance problem does not show up right away, but takes a few days to appear. When it gets bad, we see exactly the same behavior when looking at `iotop`, where any processes doing more than trivial I/O show as 99.9% IOWAIT time. The system performs sluggishly (notably the container runtime) and a reboot "fixes" the problem. Our strategy at the moment is roll back to the -136 kernel on all of our bare metal Kubernetes nodes, and hopefully get upgraded to Ubuntu 20.04 by the end of the year. If any further testing is requested please let us know and we can try things out on our dev systems. -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1928744 Title: Disk IO very slow on kernel 4.15.0-142-generic To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1928744/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs