Skip to content

UCX: applications do not work without charmrun/mpirun #2477

@matthiasdiener

Description

@matthiasdiener

Crash observed on application startup on golub:

charm/tests/charm++/simplearrayhello $ ./hello
------- Partition 0 Processor 0 Exiting: Called CmiAbort ------
Reason: UCX: UcxInitEps: runtime_kvs_put error failed: 5
[0] Stack Traceback:
  [0:0] _Z14CmiAbortHelperPKcS0_S0_ii+0x4d  [0x54800d]
  [0:1]   [0x54812d]
  [0:2] _Z8LrtsInitPiPPPcS_S_+0xd6f  [0x54d2bf]
  [0:3] ConverseInit+0x1ce  [0x54d8de]
  [0:4] charm_main+0x27  [0x48c307]
  [0:5] __libc_start_main+0xf5  [0x2abf9af0b495]
  [0:6]   [0x486360]

Running with Charmrun works fine:

charm/tests/charm++/simplearrayhello $ ./charmrun ./hello

Running on 1 processors:  ./hello
Charm++> Running in non-SMP mode: 1 processes (PEs)
Converse/Charm++ Commit ID: v6.9.0-535-g549280a73
[...]

Metadata

Metadata

Assignees

Labels

UCXThe UCX machine layer

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions