Repository navigation

feat(signal): add first-class MPIJob certification - #963

Open

Kevin Cho (chokevin) wants to merge 59 commits into

Azure:mainfrom

chokevin:chokevin-upstream-mpijob-adapter

Open

feat(signal): add first-class MPIJob certification#963
Kevin Cho (chokevin) wants to merge 59 commits into
Azure:mainfrom
chokevin:chokevin-upstream-mpijob-adapter

Commits on Oct 26, 2026
- feat(signal): add generic certification controller
  Kevin Cho (chokevin)
  committed
  a36485b
- feat(signal): support grouped node certification
  Kevin Cho (chokevin)
  committed
  11432b0
- signal: add participant backend adapter
  Kevin Cho (chokevin)
  committed
  aec313e
- signal: fix adapter formatting
  Kevin Cho (chokevin)
  committed
  c3e1d17
- signal: satisfy pinned adapter formatting
  Kevin Cho (chokevin)
  committed
  53abeee
- signal: add typed workload execution policy
  Kevin Cho (chokevin)
  committed
  2f55243
- signal: support participant host networking
  Kevin Cho (chokevin)
  committed
  82dad3f
- fix(signal): align script digest contract
  Kevin Cho (chokevin)
  committed
  93886b4
- fix(signal): request Kueue topology
  Kevin Cho (chokevin)
  committed
  366ea2f
- fix(signal): preserve diagnostic results
  Kevin Cho (chokevin)
  committed
  5ffa1af
- feat(signal): add runtime execution contract
  Kevin Cho (chokevin)
  committed
  da72b22
- fix(signal): group deadline parameters
  Kevin Cho (chokevin)
  committed
  96b2eb8
- fix(signal): guard optional immutable fields
  Kevin Cho (chokevin)
  committed
  a8b50dd
- fix(signal): preserve backend failure through cleanup
  Kevin Cho (chokevin)
  committed
  6acf82b
- fix(signal): fence MPI worker readiness
  Kevin Cho (chokevin)
  committed
  ecffc97
- fix(signal): probe MPI workers over SSH
  Kevin Cho (chokevin)
  committed
  ab56f86
- fix(signal): harden MPI launcher recovery
  Kevin Cho (chokevin)
  committed
  f53f7f6
- fix(signal): preserve pre-cleanup backend evidence
  Kevin Cho (chokevin)
  committed
  409d4a8
- docs(signal): define evidence crash boundary
  Kevin Cho (chokevin)
  committed
  cafcce5
- fix(signal): recover canonical workflow outputs
  Kevin Cho (chokevin)
  committed
  c08ebd2
- feat(signal): group certification checks in one run
  Kevin Cho (chokevin)
  committed
  b94817a
- fix(signal): resolve grouped workflow template inputs
  Kevin Cho (chokevin)
  committed
  4a196c3
- fix(signal): use valid terminal DAG dependencies
  Kevin Cho (chokevin)
  committed
  69cb1b0
- test(signal): satisfy grouped workflow lint
  Kevin Cho (chokevin)
  committed
  e25d6aa
- fix(signal): run grouped tasks with adapter service account
  Kevin Cho (chokevin)
  committed
  e810801
- refactor(signal): trim certification to production path
  Kevin Cho (chokevin)
  committed
  b7bebf2
- fix(signal): fence certification cleanup and validate execution evidence
  Kevin Cho (chokevin)
  committed
  3d85dde
- fix(signal): default backend references to the run namespace
  Kevin Cho (chokevin)
  committed
  9100fa3
- fix(signal): preserve standalone certification semantics
  Kevin Cho (chokevin)
  committed
  18f1b69
- fix(signal): guard one-time certification invariants
  Kevin Cho (chokevin)
  committed
  1eaa761
- fix(signal): preserve certification audit state
  Kevin Cho (chokevin)
  committed
  2d34a6e
- refactor(signal): narrow certification contract
  Kevin Cho (chokevin)
  committed
  9a2a1b7
- fix(signal): harden one-time certification lifecycle
  Kevin Cho (chokevin)
  committed
  a6cc036
- fix(signal): satisfy lifecycle lint checks
  Kevin Cho (chokevin)
  committed
  f33e89b
- fix(signal): align lint spacing
  Kevin Cho (chokevin)
  committed
  dae1bbb
- fix(signal): separate cancellation evidence steps
  Kevin Cho (chokevin)
  committed
  76e30eb
- refactor(signal): snapshot current script contents
  Kevin Cho (chokevin)
  committed
  076cdc1
- refactor(signal): certify one participant group
  Kevin Cho (chokevin)
  committed
  74a2609
- fix(signal): separate participant encoding
  Kevin Cho (chokevin)
  committed
  fb23ece
- refactor(signal): minimize certification check contract
  Kevin Cho (chokevin)
  committed
  d7faf3e
- feat(signal): add baseline runner image
  Kevin Cho (chokevin)
  committed
  6d79e31
- ci(signal): validate runner image
  Kevin Cho (chokevin)
  committed
  6b18dbf
- fix(signal): remove stale JSON decoder
  Kevin Cho (chokevin)
  committed
  d5c188e
- style(signal): format adapter cleanup
  Kevin Cho (chokevin)
  committed
  221da73
- refactor(signal): split runner image into separate PR
  Kevin Cho (chokevin)
  committed
  feee8be
- fix(signal): enforce executable participant limit
  Kevin Cho (chokevin)
  committed
  21cc18d
- refactor(signal): make runner deployment-owned
  Kevin Cho (chokevin)
  committed
  5406064
- fix(signal): coordinate MPI checks across participants
  Kevin Cho (chokevin)
  committed
  0fe2b0c
- style(signal): satisfy coordinated MPI lint
  Kevin Cho (chokevin)
  committed
  9bc13e5
- style(signal): separate result expansion loop
  Kevin Cho (chokevin)
  committed
  12a1455
Commits on Oct 27, 2026
- fix(signal): harden deployment contracts
  Kevin Cho (chokevin)
  committed
  8d013d3
- refactor(signal): minimize certification contracts
  Kevin Cho (chokevin)
  committed
  957c9c7
- style(signal): satisfy contract cleanup lint
  Kevin Cho (chokevin)
  committed
  7629ebc
- feat(signal): pin immutable script bundles
  Kevin Cho (chokevin)
  committed
  ec57145
- style(signal): satisfy bundle digest lint
  Kevin Cho (chokevin)
  committed
  18b6d17
Commits on Oct 28, 2026
- fix(signal): mount adapter service account token
  Kevin Cho (chokevin)
  committed
  b9fe9eb
- fix(signal): quiesce retained MPI workers
  Kevin Cho (chokevin)
  committed
  0d5413b
- style(signal): satisfy MPI cleanup lint
  Kevin Cho (chokevin)
  committed
  e9eb9d1
- fix(signal): render valid multi-node MPI affinity
  Kevin Cho (chokevin)
  committed
  7741ef0

Commits

Commits on Oct 26, 2026

Commits on Oct 27, 2026

Commits on Oct 28, 2026