IFCA Datacenter usage guidelines

1. Introduction

The GRIDUI (Grid User Interface) cluster is the interactive gateway to the Advanced Computing and e-Science resources at IFCA. This cluster is formed by several identical Scientific Linux 5 hosts, that can be reached through a single entry point. The connections to the internal machines are managed by a director node that tries to ensure that proper balancing is made across the available nodes at a given moment. Nevertheless, direct access to a particular node can be obtained.

Please note that this cluster is not intended for the execution of CPU intensive tasks, for this purpose use any of the available computing resources. Every process spawned are limited to a maximum CPU time of 2 hours.

Login on these machines is provided via Secure Shell. The RSA key fingerprint of them is 46:85:91:c1:eb:61:55:34:25:2c:d6:0a:08:22:1f:77.

Outgoing SSH connections are not allowed by default from this cluster. Inactive SSH sessions will be closed after 12h.

2. Authentication

Authentication is centralized via secured LDAP. All the changes made to a user account in one node take immediate effect in the whole cluster. There is also a secured web interface, that allows a user to change his/her details, available at https://cerbero.ifca.es/. If you need to reset your account password, please contact the system administrators.

With this username you should be able to access also the ticketing system at http://support.ifca.es/

3. Grid resources

As the several virtual organizations supported use its own set of resources, it is commonly required to set up the correct environment variables in order to use the grid tools. Environment scripts are located under /nfs4/usr/etc/env/:

$ source /nfs4/usr/etc/env/ibergrid-env.sh

4. PBS Cluster

PBS cluster was decommissioned in December 2009. Please use the new SGE Cluster instead.

5. SGE Cluster

The SGE Cluster is based on Scientific Linux CERN SLC release 5.5 machines, running on x86_64. The exact number or resources available is shown on the monitorization page.

Local submission is allowed for certain projects. As stated below, there are some shared areas that can be accessed from the computing nodes. The underlying batch system is Sun Grid Engine 6.2u5. Please note that the syntax for job submission (qsub) and monitoring (qstat) is similar to the one you might be accustomed to from PBS, but there are important differences. Refer to the following sources for information:

• Using Grid Engine official documentation.

• Some useful HOWTOS.

• SGE basic usage.

For the shake of clarity, in the following examples we will only show the options relevant to that example. However, you should take into account that there are some options that are mandatory (for example, the project submission).

5.1. Job submission, projects and queues

Users should submit their jobs directly to their project by using qsub -P <project>.

$ qsub -P <project> <jobfile>

You can get the list of projects you are allowed to use by issuing:

$ /nfs4/usr/bin/rep-sge-prjs.py

A scratch area is defined for every job as the environment variable $TMPDIR. This area is cleared after the job has exited.

5.2. Notifications

Users can get email notifications whenever a job changes its status. A valid email address must be specified, using the -M option, as well as the kind of notification using the -m option. Valid values are:

• b: Mail is sent at the beginning of the job.

• e: Mail is sent at the end of the job.

• a: Mail is sent when the job is aborted or rescheduled.

• s: Mail is sent when the job is suspended.

• n: No mail is sent.

Please note thas -m can be specified several times, so the following job sumission

• $ qsub -m b -m e -M user@example.org <jobfile>

will produce an email when the job has started and ended.

5.3. Specifying resources

In order to get your submitted jobs executed fast (and to benefit from any possible backfilling), you should tweak some of the resources that your job is requesting. The more accurate you are to your job bounds, the faster your job will run (the default values are quite high in order to prevent jobs to be killed by the batch system, thus they penalize a lot the job execution).

Some of these limits are defined in the $SGE_ROOT/default/common/sge_request file. If your application is always expected to use the same values, you can override that file by creating a $HOME/.sge_request file. For further details, please check the sge_request manual page.

5.3.1. Wall Clock time

A default wall clock time of 72 hours is enforced by default in all jobs submitted to the cluster. Should you require a higher value, please set it by yourself by requesting a new h_rt value in the form hours:minutes:seconds. Please note that requesting a high value may impact negatively in your job scheduling and execution. Please try to be as accurate as possible when setting this required value. For example, a job requiring 22h, with a couple of "safety" one extra hours should be sent as follows:

$ qsub -l h_rt=24:00:00 <jobfile>

5.3.2. Memory management

When requesting memory for a job you must take into account that per-job memory is limited in the default queues to a Resident Set Size (h_rss) of 5 GB. If you need to use more memory, you should request the special resource highmem. Please notice that your group may not be able to request that flag by default. If you need to do so, please open a ticket requesting it. Also, notice that these nodes might be overloaded by other users requesting the same flag, so use it wisely.

It is highly recommended that you tune your memory requirements to some realistic values. Special emphasis is made in the following resources:

• h_rss

• mem_free

5.3.2.1. h_rss

This limit refers to the hard resident set size limit. The batch system will make sure a given job does not consume more memory than the value assigned to that variable. This means that any job above the requested h_rss limit will be killed (SIGKILL) by the batch system. It is recommended to request this resource as a top limit for your application. If you expect your job to consume no more than a peak value of 3GB you should request those 3GB as its resident set size limit. This request shall not produce a penalty on the scheduling of your jobs.

5.3.2.2. mem_free

This refers to the free RAM necessary for the job to run. The batch system will allow jobs to run only if sufficient memory (as requested by mem_free) is available for them on a given node. It will subtract that amount of memory from the available resources, once the job is running. This ensures that a node with 16 GB of memory will not run jobs totaling more than 16 GB. The default value is 1.8 GB per slot. Please note that breaking the mem_free limit will not automatically kill your job. Its aim is just to try to ensure that your job has available the memory you requested. Also note that this value is not intended to be use to reflect the memory peaks of your job. This request will impact the scheduling of your jobs, so it is highly recommended to tune it to fit your actual application memory usage.

5.3.2.3. Memory usage above 5G

For serial jobs requiring more than 5 GB of memory, submission requesting the highmem flag is necessary. Using this flag, the h_rss limit will be unset, but the requirement tuning described above still applies. If your group is allowed to request it, and your job needs 20GB of memory, you can request it as follows:

$ qsub -l highmem,mem_free=20G <jobfile>

• A job that might reach 30 GB and needs 20GB will be submitted as:

• $ qsub -l highmem,h_rss=30G,mem_free=20G <jobfile>
  

For jobs needing more than these 5G using MPI, please refer to the Parallel job submission section.

5.3.2.4. Examples

• A job that needs to have 4 GB of memory assigned to it:

• $ qsub -l mem_free=4G <jobfile>
  

• A job that might peak at 4 GB, but in its execution normally needs 3 GB:

• $ qsub -l h_rss=4G,mem_free=3G <jobfile>
  

• A job that might reach 4 GB, and also needs 4 GB:

• $ qsub -l h_rss=4G,mem_free=4G <jobfile>
  

5.3.3. Infiniband

If you are executing MPI parallel jobs you may benefit from the Infiniband interconnection available on the nodes. In order to do so, you must request the special resource infiniband:

$ qsub -l infiniband <jobfile>

5.4. Job reservation

It is possible to indicate whether a reservation for a job should be done or not using the -R y option. When a runnable job can not be started due to a shortage of resources a reservation can be scheduled instead. This is specially useful for parallel jobs, or jobs requesting bottleneck resources (for example, a high amount of memory), and it is useless for normal sequential jobs.

$ qsub -R y <jobfile>

5.5. Parallel jobs

Parallel jobs must be submitted to a parallel environment (pe), specifying the number of slots required. Depending on the used pe, SGE will allocate the slots in a different way.

$ qsub -pe mpi 8 <jobfile>

Please note that parallel jobs will be routed to queue parallel (see previous section). Also note that access to that queue is restricted to groups having requested it beforehand.

The following parallel environments are available:

The actual command line for the execution of the parallel application depends on the parallel library/framework that it uses for communication between the allocated slots. At IFCA, Open MPI v.1.4 is configured in the default user environment and its installed at /usr/lib64/openmpi/1.4-gcc.

5.6. Interactive jobs

Interactive, short lived and high priority jobs can be sent if your project has permission to do so (see the SUBMIT manual page (man submit)). This kind of jobs can only request a maximum of 1h of WALL clock time, see previous section for details about limiting the wall clock time of a job.

X11 forwarding is possible when using the qlogin command. Using X11 forwarding requires a valid DISPLAY, use ssh -X or ssh -Y to enable X11 forwarding in your ssh session when logging in the UI.

$ qlogin -P <project> -l h_rt=1:00:00

5.7. Pseudo-Interactive jobs

A special resource, called immediate is available for some users, that need fast scheduling for their short-lived batch jobs. This kind of jobs can only request a maximum of 1h of WALL clock time.

$ qsub -l immediate <jobfile>

Please note that you might not have access to these resources.

5.8. Resource quotas

Some limits may be enforced by the administrators in a user/group/project basis. To check the current resource quotas, the following command must be issued:

$ qconf -srqs

In order to know the current usage of the quotas defined above, the comand qquota must be used:

$ qquota -P <project>

5.9. Advanced reservation

Some users and/or projects might request a reservation of a set of resources in advance. This is called an "Advanced Reservation (AR). If your project needs such a reservation you should make a petition using the support helpdesk. You need to specify the following:

• Start datetime, and end datetime (or duration) of the reservation.

• Duration of your job(s) (i.e. h_rt for the individual jobs).

• Computational resources needed (mem_free, number of slots).

Once the request has been made, the system administrators will give you the ID(s) of the AR created. You can submit your jobs whenever you want by issuing:

$ qsub -ar <reservation_id> <other_job_options>

You can submit your job(s) before the AR starts and also once it is started. However, you should take care of the duration of the reservation and your job' duration. If your job execution exceeds either the h_rt that it has requested or the duration of the AR it will be killed by the batch system.

You should also take into account that your reservation might not be created in the date and time that you requested if there are no resources available. In this case, it will be created whenever it is possible. To avoid this, please request your reservations well in advance.

Since the requested and reserved resources cannot be used for other jobs, those requested resources will be used for accounting purposes as if they were resources used by normal jobs (even in the case that the AR is unused). Please request only the resources that you need.

If you want to query the existing advance reservations, you can use the qrstat command. To query about an specific advance reservation, you can issue:

$ qrstat -ar ''<reservation_id>''

6. Shared areas

The $HOME directories (/home/$USER) are shared between the UIs and the computing nodes. There is a projects shared area (located at /gpfs/csic_projects/), also accessible from the UI and the computing nodes. If your group does not have this area, please open an Incidence ticket.

6.1. Usage

The shared directories are not intended for scratch, use the temporal areas of the local filesystems instead. In other words, instruct every job you send to copy the input from the shared directory to the local scratch ($TMPDIR), execute all operations there, then copy the output back to some shared area where you will be able to retrieve it comfortably from the UI.

As mentioned above, the contents of $TMPDIR are removed after job execution.

6.2. Disk quotas

Disk quotas are enabled on both user and projects filesystems. A message with this information should be shown upon login. If you need more quota on your user space (not in the project shared area), please contact the system administrators explaining your reasons.

If you wish to check your quota at a later time, you can use the commands mmlsquota gpfs_csic (for user quotas) and mmlsquota -g id -g gpfs_projects (for group quotas). A script reporting both quotas is located on /nfs4/usr/bin/rep-user-quotas.py. A sample output of the latter could be:

**********************************************************************
                    INFORMATION ABOUT YOUR CURRENT DISK USAGE

USER                Used      Soft      Hard     Doubt     Grace
Space (GB):         3.41     20.00      0.00      0.06      none
Files (x1000):        64         0         0         0      none

GROUP               Used      Soft      Hard     Doubt     Grace
Space (GB):         0.00   1000.00   1500.00      0.00      none
Files (x1000):         0         0         0         0      none
**********************************************************************

For a basic interpretation of this output, note that the "Used" column will tell you about how much disk space you are using, whereas "Soft" denotes the limit this "Used" amount should not exceed. The "Hard" column is the value of the limit "Used" plus "Doubt" should not cross. A healthy disk space management would require that you periodically delete unused files in your $HOME directory, keeping its usage below the limits at all times. In the event that the user exceeds a limit, a grace period will be shown in the "Grace" column. If the user does not correct the situation within the grace period, she will be banned from writing to the disk.

For further information you can read the mmlsquota command manual page.

7. Extra utils/Software

Some extra packages as latest Python versions and Intel Non-Commercial Compilers can be found at /nfs4/opt/. Here also is the preferred location for some other piece of software commonly used like:

• Matlab's like octave language for numerical anaylisis.

• Data plotting gnuplot program.

• Profiling and debugging valngrid tools.

Please note that these packages are provided as-is, without further support from IFCA staff.

8. Support

Before opening a new incidence, please check the Frequently Asked Questions page

Questions, support and/or feedback should be directed through the use the Helpdesk.

CategoryUserSupport CategoryLocalCluster