vrun.ggo

Path MPICH-V for MPICH the MPI Implementation

package "vrun"
purpose "Launches and monitors an MPI application in the V-environment"
option "n-procs"              n "number of processes for the MPI application (np flag of mpirun)" int typestr="np" yes
option "jobid"                j "unique id for the job. (default=rand)" int typestr="int" no
option "machines-file"        m "machines file like for any MPICH device (typical deployment may use at least 3 more machines than np)" string typestr="filename" default="machines" no
option "stable-machines-file" - "file of stable machines (machines having a high MTBF)" string typestr="filename" no
option "n-el"                 e "number of event loggers" int typestr="int" default="1" no
option "n-cs"                 c "number of checkpoint servers (default=np/5)" int typestr="int" no
option "runtime-port"         - "port for the runtime process (default=autoselect)" short typestr="port" no
option "stdio-port"           - "port for the collection of standard Input/Output of servers and processes" short typestr="port" no
option "stder-port"           - "port for the collection of error Output of servers and processes" short typestr="port" no
option "el-port"              - "port for the event logger (default=autoselect)" short typestr="port" no
option "cs-port"              - "port for the checkpoint server (default=autoselect)" short typestr="port" no
option "ckpt-use-local-copy"  - "set if the checkpoint should use a local copy (improves fault recovery time in global coherent checkpoint protocols)" flag off
option "ckpt-timeout"         - "timeout of checkpoints (0 for no checkpoint)" int typestr="seconds" no
option "ckpt-strategy"        - "strategy of checkpoints. Use for Message-logging protocols" string typestr="(nockpt|random|roundrobin)" default="roundrobin" no
option "el-strategy"          - "strategy of event logger. Use for Message-logging protocols with event logger (only for performance measurement)" string typestr="(normal|slow)" default="normal" no
option "sc-port"              - "port for the checkpoint scheduler (default=autoselect)" short typestr="port" no
option "rsh"                  - "remote shell program for executing remote commands" string typestr="command" default="/usr/bin/ssh" no
option "working-dir"          - "working directory for running the MPI application (default=current pwd)" string typestr="path" no
option "tmp-dir"              - "writeable directory (this directory must be writeable on all machines of the machines file and stable machines file. A local filesystem is necessary for performances)" string typestr="path" default="/tmp" no
option "helpers-dir"          - "directory with the helpers (checkpointserver, etc...) (default=same directory as vrun)" string typestr="path" no
option "master-sched"         - "address of the master scheduler (only for multiple applications scheduling)" string typestr="ip:port" no
option "debug"                d "debugging options. First part is the target, second part the selection. Target may be %filename for storing in a file, or ip:[port] for sending debugs to a debug server. Selection use +label or -label to force or prevent the displaying of certain informations." string typestr="[%filename|ip:[port]][+|-labels]*" no