Repository navigation
Nitrum 0.2.1
Reliability and diagnosability patch: crash-loop detection in the control-plane, fatal-error telemetry that actually flushes, clearer KMS AccessDenied diagnostics, and a KMS permissions guide.
Highlights
- Crash-loop detection — control-plane watches
describe-enclavesafterrun-enclavefor a settle window instead of a fixed warmup sleep; an early disappearance is reported asexited_before_stableand does not reset backoff - Fatal telemetry flush — data-plane and control-plane install a stdout tracing subscriber before networking/IMDS/SSM and return
ExitCode(noprocess::exit), so exporters flush on exit and full error chains are logged - KMS diagnostics — attested KMS Decrypt
AccessDeniedis logged once per boot with the key id, ImageSha384 / PCR0 and optional NSM PCR0 - Docs — new KMS permissions and troubleshooting guide (docs/kms.md)
Security
h20.4 updated to a patched release (RUSTSEC-2026-0258) andrustls0.23 refreshedcargo-denyignores RUSTSEC-2026-0258 forh20.3, which the AWS SDK still pulls in and which has no upstream patch
Upgrading
nitrum init pins the runtime images to @sha256: digests in nitrum.toml, so existing projects keep running the 0.2.0 control-plane and data-plane until you re-pin [runtime] control_plane and data_plane to the v0.2.1 images (by tag or new digest).
Changing data_plane changes the EIF: run nitrum build and nitrum cloud deploy again. This updates PCR0 in the KMS key policy as usual (see docs/kms.md).
No breaking changes.
Install
curl -fsSL https://raw.githubusercontent.com/matzapata/nitrum/v0.2.1/scripts/install-nitrum.sh | NITRUM_VERSION=v0.2.1 shFull changelog: v0.2.0...v0.2.1