linux-system-roles / kernel_settings Goto Github PK

Below is a snippet from a correspondence with Joe Mario regarding which sysctl fields from his area of expertise should be added to the kernel_report module:

as for my area (perf/bpf) I can tell you following parameters make some
sense to move and should be just simple read/write to do so:

kernel.perf_cpu_time_max_percent
kernel.perf_event_max_contexts_per_stack
kernel.perf_event_max_stack
kernel.perf_event_mlock_kb
kernel.perf_event_paranoid

kernel.bpf_stats_enabled
kernel.unprivileged_bpf_disabled
net.core.bpf_jit_enable
net.core.bpf_jit_harden
net.core.bpf_jit_kallsyms
net.core.bpf_jit_limit

These should be added to the list of acceptable sysctl fields and tested at some point in the future. Tied to PR #58.

I believe there is a bug in here,

https://github.com/linux-system-roles/kernel_settings/blob/master/tasks/main.yml#L115

This line I believe should not be in the when clause,

- not kernel_settings_reboot_ok | d(false)

or should be

- kernel_settings_reboot_ok | d(false)

otherwise, the handler will never fire

The README states that tuned is the default provider for Fedora/RHEL, this is reasonable.

However, it looks like there are currently no other providers available, that might be worth pointing out. This also means using the role to operate on sysctl files is not possible. That might be helpful in cases where there are other applications and automation in play which deal only with sysctl files for kernel settings so from admin/user perspective keeping everything under /etc/sysctl.d would perhaps be more consistent.

I would consider this as a relatively low-priority RFE at this point. Thanks.