Kura 0.10.0 improves how the artifact cache sizes itself on disk, so deployments stop running out of cache space too soon while still avoiding disk-full failures.
- The CAS segment-ring capacity is no longer fixed at ~2.5 GiB. It now derives from the data-dir filesystem size by default: the budget is set to half the filesystem, capped at 80% of the filesystem so the cache cannot grow large enough to cause an out-of-space error.
- You can override the derived budget with the new
KURA_CAS_CAPACITY_BYTESenvironment variable, or via the Helm chart valueconfig.casCapacityBytes. Any explicit value is also clamped to the 80% filesystem cap. - Tiny or unknown-size volumes fall back to the legacy 1/2/2 segment ring, so small deployments keep the same behavior as before.
- The change is node-local only: there is no on-disk format, wire format, or replication protocol change, so mixed-version meshes continue to work safely.