使用MPI在多个节点上启动ipyparallel集群

San*_*eep 5 cluster-computing mpi ipython ipython-parallel

我正在尝试使用MPI启动ipyparallel集群.

ipcluster_config具有如下修改的以下行:

c.MPILauncher.mpi_cmd = ['mpiexec']
c.MPIControllerLauncher.controller_args = ['--ip=*']
c.MPILauncher.mpi_args = ["-machinefile", "~/mpi_hosts"]
Run Code Online (Sandbox Code Playgroud)

ipcontroller_config.py配置如下:

c.HubFactory.engine_ip = '*'
c.HubFactory.ip = '*'
c.HubFactory.client_ip = '*'
Run Code Online (Sandbox Code Playgroud)

但是,当我使用命令启动集群时, ipcluster start --profile mpi -n 2 它会失败并显示以下消息

Engines shutdown early, they probably failed to connect.
You can set this by adding "--ip='*'" to your ControllerLauncher.controller_args
Run Code Online (Sandbox Code Playgroud)

不确定如何进一步调试.