I'm looking to enable Workload Management for Splunk and I'm just trying to understand whether this is fully supported on a deployment which is using cgroups v2.
Based on the documentation, the wording on the Configure Linux systemd for workload management page suggests that Splunk understands cgroups v1:
CPU: /sys/fs/cgroup/cpu/system.slice/<SPLUNK_SERVER_NAME>.service Memory: /sys/fs/cgroup/memory/system.slice/<SPLUNK_SERVER_NAME>.service
But there is nothing (except the diagram) that suggests it can operate under cgroups v2 as well.
Does anyone know whether Splunk fully supports cgroups v2, or will our deployment need to be cgroups v1?
Edit: Some investigation suggests that Splunk fails to start with Workload Management enabled on a cgroups v2 host:
Jul 15 10:30:50 hostnamehere splunk: Couldn't open dir /sys/fs/cgroup/cpu/: No such file or directory Jul 15 10:30:50 hostnamehere splunk: Error perform systemd preparation: No such file or directory
This is the same thing I needed to do for RHEL 9.
I just encountered this issue with Fedora 33
My fix was to edit /etc/systemd/system/Splunkd.service then change the ExecStartPost commands to
# old cgroups v1 method installed by Splunk. Comment these out # ExecStartPost=/bin/bash -c "chown -R splunk:splunk /sys/fs/cgroup/cpu/system.slice/%n" # ExecStartPost=/bin/bash -c "chown -R splunk:splunk /sys/fs/cgroup/memory/system.slice/%n" # new cgroups v2 method ExecStartPost=/bin/bash -c "chown -R splunk:splunk /sys/fs/cgroup/system.slice/%n"
Don't forget to run "systemctl daemon-reload" after changing the file.
Edit: Not sure if there are other implications with cgroups v2 as I haven't tested workload management but at least Splunk starts with that fix.