Make Idunn admission atomic and request-bound

This commit is contained in:
Meta Vens 2026-07-17 02:37:42 +01:00
commit 92fab01d37
10 changed files with 1472 additions and 112 deletions

View file

@ -110,8 +110,10 @@ The current typed records are:
```text
idunn.desired_daemon.v1
idunn.daemon_health.v1
idunn.signed_health_admission.v1
idunn.keepalive_decision.v1
idunn.deployment_request.v1
idunn.deployment_request.v2
idunn.current_deployment_request.v1
idunn.deployment_result.v1
idunn.release_target.v1
idunn.deployment_artifact.v1
@ -184,9 +186,12 @@ idunn.rudp_health_ingress.v1
and then publishes `idunn.rudp_health_ingress.v1`. The Starfire local
supervisor binds `0.0.0.0:17870` explicitly so host-local and WireGuard
publishers can use the configured Idunn health endpoint. That ingress
accepts only raw `idunn.daemon_health` CultNet document puts on the `schema`
channel, decodes the typed MessagePack payload, and writes it into the
keepalive store. Each one-shot publisher gets its own RUDP session from its
accepts raw `idunn.daemon_health` CultNet document puts and the dedicated
signed Epiphany runtime-health schema on the `schema` channel. Generic health
is observation-only. Epiphany promotion requires the separately persisted
`idunn.signed_health_admission.v1`, an exact latest
`idunn.deployment_request.v2`, a pinned host signature, current matching
health, and a fresh active admission. Each one-shot publisher gets its own RUDP session from its
UDP source address and the session is discarded after a delivered health
frame. Windows UDP `ConnectionReset`/`ConnectionAborted` reports from closed
one-shot clients are nonfatal ingress noise, not a reason to kill the worker.
@ -285,7 +290,7 @@ idunn.rudp_health_ingress.v1
execution. Operators publish `idunn.lifecycle_command.v1` with `idunn restart
--daemon <id>` or `idunn redeploy --daemon <id>`. The running supervisor
claims pending records for its target catalog, writes the corresponding
`idunn.restart_request.v1` or `idunn.deployment_request.v1`, executes through
`idunn.restart_request.v1` or `idunn.deployment_request.v2`, executes through
the same guarded actuator path used by health-driven decisions, and records
the matching result document. If a target lacks restart or deploy authority,
Idunn marks the lifecycle command rejected instead of inventing authority.