WeeklyTelcon_20170404

Open MPI Weekly Telcon

---

• Dialup Info: (Do not post to public mailing list or public wiki)

Attendees

• Josh Hursey (IBM)
• Howard (LANL)
• Artem Polyakov (Mellanox)
• Joshua Ladd (Mellanox)
• Edgar Gabriel (UH)
• Todd Kordenbrock (Sandia)
• Ralph (Intel)
• Sylvain Jeaugey (NVIDIA)
• Brian Barrett (Amazon)
• David Bernholdt (ORNL)
• Geoffroy Vallee (ORNL)
• Thomas Naughton (ORNL)
• George (UTK)

---

Agenda

Review All Open Blockers

• Issue #3267 - OSU crashing in v2.x series
  • PR #3274 Needs to be PR'ed to v2.x series
  • Issue should be re-opened until it has been PR'ed to all of the correct places.

Review v2.0.x

• No specific update

Review v2.1.x

• PR #3193 - RM's need to review and discuss a bit more.
• PR for Issue #3267 needs to be PR'ed to the v2.x series
• Issue #3268 -

Review v3.0.0

• PR #3271 - Update hwloc to 1.11.6
  • Pushed to the next release
  • Add a release note about this issue, and possible workarounds (e.g., external hwloc)
• Disabling CMA
  • PR #3272 vs PR #3270 : Only one of these should be picked

Review PMIx v2.0

• v2.0 API's are all in master - we are API complete
  • Open MPI master has been updated (PR #3273)
• Timeline:
  • Target PMIx v2.0 release end of April
  • Commit this to master, once stable MTT, then PR to v3.0

Review Master Pull Requests

• No update

---

AWS Testing Setup

• Jenkins setup in progress
  • Jenkins Builder "Open MPI CI"
    • Builds after every push to ompi (not PRs yet)
    • Basic build running at UH
    • Infrastructure coming along
• MTT Setup
  • Will test using many of the open-source batch schedulers (e.g., SLURM, Torque)

Review Master MTT testing

• Need more participation in MTT testing
  • In particular those pushing for a faster release cycle
  • IBM: To turn this back on this week.
  • Mellanox: To start getting this setup soon.
• Absoft has a compiler failure
  • Ralph to reach out to check on that
• Cisco:
  • Disabled one-side tests
  • oshmem still failing

MTT Development status:

• No update

---

Exceptional topics

• If you have a patch that needs to go to a release branch
  • Label the PR to master with the Target X tag so it doesn't get missed
  • Do not close the Issue until the patch has been PR'ed and merged into all of the required release branches.
    • Be careful of auto-closing issues from your PR commit message.
• Next Face-to-Face meeting
  • Doodle poll

---

Status Updates:

• Mellanox
  • PMIx, v2.x ongoing work
  • Working on adding a osc/ucx component
• Sandia
  • Issue #173 - Very old ticket on Portals MTL
  • MTT - Trying to figure out why some tests are running very slowly
• Intel
  • PMIx v2.0
    • Cross version work
    • Solidifying master\v2.x branch
  • OpenMP/MPI coordination
  • Launch scaling work

Status Update Rotation

1. Mellanox, Sandia, Intel
2. LANL, Houston, IBM, Fujitsu
3. Cisco, IBM, ORNL, UTK, NVIDIA, Amazon

---

Back to 2017 WeeklyTelcon-2017