benpope

This is exactly the reason for the default 10% buffer between what we tell seastar it can have, and what we request for the cgroup with K8s.

In one case we saw that the kernel was unable to allocate a TCP buffer, so it decided to OOMKill Redpanda.

This should work fine. If you have any issues, please report as a bug.

The tuning isn't 100% necessary, it's there primarily for running max performance on metal, and you can pick and choose what to tune.

In general it probably best to run with:

--smp <n> --overprovisioned

for a container with <n> CPUs.

These are standard Seastar flags.

This is not a hard requirement, but XFS is highly recommended for production.