- Shipped
- September 24, 2026 at 10:43 PM UTC
- Author
- Kamo
- Commit
- b08115b
Seastar keeps max(1.5 GiB, 7%) of the cgroup limit plus ~94 MiB per shard back for the OS and refuses to start when --memory does not fit in what is left. With the default reserve a 4Gi pod leaves 2.32 GiB, so --memory 3G exited at once ("insufficient physical memory: needed 3116367872 available 2487222272") and scylladb-0 never became Ready. --reserve-memory 768M leaves 3.07 GiB. A throwaway instance started in the pod's own cgroup failed with the old arguments and reached the reactor with the new ones.