General usage of jumel and juman

Quick introduction

Before you can use the logger you need to load the corresponding module. The available versions of jumel can be listed on JUROPA with

module av jumel

module load jumel

In order to get an overview of the valid options, please use

jumel -u
juman -u

To explain the general usage, suppose the application to monitor is started usually as follows:

mpiexec -np 32 -e APP_ROOT app.x -i app.inp > my.out

To start the application with jumel just use

mpiexec -np 32 -e PBS_ID,APP_ROOT\
 jumel -n -a "app.x -i app.inp" > my.out

The variable PBS_ID is not needed, however if specified it will be displayed in the jumel logfiles and eases tracking runs afterwards (e.g. when looking for the job in the system logfiles).

Once the run is finished run juman in the same directory:

juman -s all -n -i

The -s option switches on the statistics and the -i option will start the graphical display of the results. Currently only a gnuplot interface is implemented and postscript or xfig files can be generated. An interface for visualization with Python is planned. Gnuplot, gv (needed for viewing postscript files with juman) and xfig are not part of the jumel software.

Options of jumel

Currently, the following options are supported by jumel (use jumel -u to display this list on a terminal):

 -a, --application "application"
 Application to be logged. Must be given in ""
 Option is MANDATORY

 -d, --delay <int seconds>
 Define the interval for sampling the memory
 (in seconds).
 Default value is 10

 -f, --flush <int>
 Defines how the ouput is flushed:
 0 : No buffering/immediate flushing
 1 : Buffering linewise
 n : Buffering approx. n bytes
 -1 : System default
 The default is 0

 -h, --help
 Print short help information

 -n, --node
 Log resource consumption for each node. The
 process running on core 0 will perform
 the logging and save the data in
 wrkdir/nodeXXXX_taskYYYY.log.

 -o, --offset <int seconds>
 Time to wait before starting to log
 (in seconds).
 Default value is 0

 -p, --print_resources
 Adds a table with resources available on each
 node to the file
 wrkdir/nodeXXXX_taskYYYY.log

 -r, --rank <int rank>
 Monitor only the MPI rank specified.
 Default is -1 (monitor all ranks).

 -t, --task
 Log resource consumption for each task.
 If -r is specified this will be done only for
 the task requested, otherwise for all tasks.
 Results will be safed in
 wrkdir/taskYYYY.log.

 -u, --usage
 Print this extended help

 -w, --wrkdir <char directory>
 Directory for the log files of the experiment
 The default is .memlog

 -v, --verbose
 Verbose output written to STDERR

 -V, --version
 Prints the version number and exists

Options of juman

Currently, the following options are supported by juman (use juman -u to display this list on a terminal):

 -f, --format <char format>
 Define the graphics format.
 Currently supported formats:
 ps : postscript (.ps)
 fig : xfig (.fig)
 The default is ps

 -h, --help
 Print short help information

 -i, --imageview
 Displays generated graphs/images

 -n, --node
 Perform analysis for node data
 Data in wrkdir/nodeXXXX_taskYYYY.log are
 considered.

 --nodekey <char key>
 Specifies the key word for the node analysis.
 The default is MFree

 -p, --plot <char plottool>
 Specifies the tool to generate the graphs.
 Currently supported:
 g : gnuplot
 The default is g

 -s, --statistics <char statistics>
 Perform statistical analysis on the data.
 Currently supported:
 None : Perform no statistical analysis
 minmax : Maximum and minimum value of key
 per step across all tasks
 nodes : Print key values for nodes.
 tasks : Print key values for tasks.
 total : Sum of key values over all tasks
 per step
 all : Complete statistics
 The default is minmax

 -t, --task
 Perform analysis for task data.
 Data in wrkdir/taskYYYY.log are considered.

 --taskkey <char key>
 Specifies the key word for the task analysis.
 The default is VmSize

 -u, --usage
 Print this extended help

 -V, --version
 Prints the version number and exists

 -w, --wrkdir <char directory>
 Directory with the log files of the experiment
 The default is .memlog

 --watch
 Update statistics every <watchinc> seconds.
 The default is no update

 --watchinc <float>
 Sets the interval n seconds for updating
 statistics
 The default is 20