It is not recommended to run very long or parallel jobs as interactive jobs in Hippu. The maximum run time of interactive jobs in Vuori is only 4 hours. Larger jobs can be submitted to Vuori compute nodes either from the Maestro GUI or directly from the command line.

Passwordless ssh within Vuori

Maestro jobs communicate with the login node with passwordless ssh. You must enable it in order to run jobs in Vuori compute nodes. To do this, copy the public key from the work directory (used as $HOME by the compute nodes) to the home directory on the login nodes:

cp $WRKDIR/.ssh/id_dsa.pub ~/.ssh/authorized_keys

In Hippu there is no queuing system, and all jobs are run on the localhost. Please run all long jobs in Vuori and leave Hippu for interactive jobs and for jobs that need exceptional amounts of memory.

GUI submission

If you have Maestro running interactively in a compute node in Vuori, you just need to click the start button after job configuration, select the queue and number of nodes, and submit the job. When the job has been submitted, you can close the Maestro GUI and open it again once the jobs are ready.

Command line submission

The Maestro jobs are not submitted with a batch job script like most other jobs. Suppose you have composed a ConfGen job with Maestro on your own computer and written it to the disk with a job name rakenteita. This will result in two files called rakenteita.mae and rakenteita.com. Copy them to your $WRKDIR in Vuori (e.g. with the Scientist's User Interface file manager). Then load the Maestro environment and submit the job as a serial job:

module load maestro
$SCHRODINGER/bmin -HOST serial rakenteita
or 
bmin -HOST serial rakenteita

-HOST flag chooses the queue to run the job. In Vuori there are two different queues (and localhost, which should be used only for serial interactive jobs in a compute node). You can get the available queues and what are their properties with

bmin -HOSTS
bmin -HOST serial -ENTRY

jobcontrol -list

bjobs

 bmin -help

A job that generates conformers using many processors simultaneously (requires that you have many molecules) can be done as follows:

$SCHRODINGER/utilities/para_bmin -NJOBS 8 -HOST serial:8 rakenteita

qsite -PROCS 8 -HOST serial qsite_input.in

A large parallel Glide docking job can be made to take advantage of the whole GlideXP license set with the following command. It splits the molecules to be docked into 180 separate jobs and runs them at most 20 at a time in the serial queue. The master job is run at the login node, but after creating the subjobs it hardly uses any resources. (The current license allows for 20 simultaneous GlideXP jobs and other stuff can be done with the floating license tokens). The input file GlideXP_EP.in created by the Maestro GUI with the write files button. Note that you also need the other input files referred to in GlideXP_EF.in

$SCHRODINGER/glide -LOCAL -strict -NJOBS 180 -HOST serial:20 GlideXP_EF.in

More examples can be found at the program specific manuals, which are available at the Scientist's Interface

A simple test job

Download here a simple test job (geometry optimization of a water molecule using MacroModel). You can also download it directly from Hippu or Vuori command line with:

wget http://www.csc.fi/english/research/sciences/chemistry/maestro_files/testi.tar

tar xf testi.tar