Description of problem: For proper isolation of CPUs, we need to tune both systemd CPUAffinity and /sys/devices/virtual/workqueue/cpumask These are both features of tuned for RHEL which does not work for OCP/RHCOS. Since systemd starts from the initramfs, it loads its config file from there, where CPUaffinity is set. See https://bugzilla.redhat.com/show_bug.cgi?id=1394965 For the kernel workqueue, this must happen before device initalization. See https://bugzilla.redhat.com/show_bug.cgi?id=1395899 The solution tuned on RHEL uses is a second ramdisk, see https://bugzilla.redhat.com/show_bug.cgi?id=1414098 We need the ability to add a second ramdisk for worker nodes to support this and potentially other early-boot customizations for tuning.
Andrew, is this BZ still relevant in the presence of: https://bugzilla.redhat.com/show_bug.cgi?id=1775917 or can this be closed?
I believe this report should be updated as NTO will likely be able to use the new systemd kernel option systemd.cpu_affinity (https://bugzilla.redhat.com/show_bug.cgi?id=1827621)
This is not a showstopper for OCP 4.5.0. Setting target release to 4.6.0 (active development branch). For fixes (if any) required/requested on earlier versions of OCP, please create clones targeting 4.5.z and earlier z-stream releases when appropriate.
This is an RFE that is being tracked internally by a Jira card. Closing.
The needinfo request[s] on this closed bug have been removed as they have been unresolved for 1000 days