Update PMIx/PRRTE submodule pointers #11563
Conversation
Includes a fix for intermittent launch hangs on main Signed-off-by: Luke Robison <[email protected]>
|
Additional testing suggests that bumping the version does not fix the intermittent hangs. Closing this PR until I find the real cause. |
|
Letting my tests run overnight does suggest some improvement on main-branch by updating the submodules. The following screenshot shows this PR labeled "Forward" on top against main as it stands now. Each dot represents an Open MPI run of a simple benchmark on 8 hosts with 64 ranks each that completed, and each F represents a failure which hangs.
As you can see the main branch as it stands now has many more failures than this PR. I will still try and find a stack trace on the remaining hangs. Some statistics: with this PR I only see 71 failures after 2370 tests (3%) |
|
Progress happening over at #11566 in finding the real root cause. Closing this PR. |
Includes a fix for intermittent launch hangs on main