Skip to content

clio-relay 1.4.13

Choose a tag to compare

View all tags
@JaimeCernuda JaimeCernuda released this 21 Jul 04:39
v1.4.13
a461b3a
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Idempotent bootstrap and recoverable session cleanup

clio-relay 1.4.13 turns cluster bootstrap into desired-state reconciliation and closes the crash window left by an interrupted owned-session teardown.

  • Reuses an exact healthy installed generation through an end-to-end 30-second gate, without building, reading, or transferring a bootstrap payload.
  • Repairs an inactive or disabled endpoint service through a separate end-to-end 60-second gate, without reinstalling JARVIS or touching its resource graph.
  • Preserves an existing JARVIS execution environment and operator-owned configuration. For a missing graph, relay asks JARVIS to activate the exact operator-selected packaged profile; only structured unavailability plus an explicit operator policy permits the no-benchmark discovery fallback.
  • Stages relay-only upgrades as crash-recoverable generations instead of clearing and rebuilding the JARVIS environment.
  • Records machine-readable bootstrap phase durations, component actions, downloads, payload bytes, service operations, queue work, and JARVIS actions.
  • Bounds queue lifetime-lock handoffs, pins migration I/O to the locked queue root, and rejects duplicate-key migration evidence.
  • Keeps POSIX queue access pinned to the verified directory descriptor, including Linux /proc/self/fd aliases, and rejects symlink, descriptor-reuse, and root-retarget races.
  • Makes owned remote-session startup and teardown resumable through pinned, owner-scoped transactions and exact process/file identities.
  • Treats a lost or timed-out session-start transport as an inspectable nonterminal operation, never as proof of failure. Operators can persist an exact operation and route selector before mutation, query it without an aggregate wait deadline, and retry only the identical request. Scheduler jobs remain separate durable executions and may stay PENDING for hours or days without becoming failed or canceled.
  • Persists and re-verifies a coordinator-finalized teardown receipt before authoritative closure, so reconnect can recover only from the same fully verified connector, gateway, relay-job, scheduler-job, and filesystem disposition.
  • Keeps full teardown evidence in bounded, crash-recoverable report artifacts while emitting compact machine-readable completion summaries and enforcing bounded artifact retention.
  • Keeps relay and scheduler jobs by default. Job cancellation and endpoint-service shutdown remain explicit opt-ins.

Runtime dependency pins are JARVIS-CD 1.4.8 and clio-kit 2.5.23.

Assets 5
Loading