PBS installation on a single execution host

Users/Site Administrators
1

Hi OpenPBS community team, I tried to install this program to manage the workload. my interest to use OpenPBS to run Numerical model simulations of WaveWatchIII. I have a single node 20 processor with GPU CUDA RX5000 configuration and OS ubuntu 20.04.1 LTS machine. So, I would like to install and run all PBS components on a single machine (Page no.13 - single execution system referred from PBS professional 2021.1 installations & Upgrade Guide). I did two trials to install but not successfully installed. First trial with Openpbs source code tar.gz file from new release v20.0.1, I have followed the instruction given in the INSTALL file but some errors(111) is showing, and the program was not running properly. The second one openpbs_20.0.1.ubuntu_1804.zip tried to install but while pbs start while i am giving terminal automatically terminated and another error showing(server .deb files are not installing properly and showing error / dpkg-deb : error: paste subprocess was killded by signal (Brocken pipe). So, Kindly help me with how to install the program

2

Hi , I did the below on a new installation of ubuntu 20.04.2 LTS and it worked without any issues

As root user:

  • set the static ip address and hostname in /etc/hosts

  • create a user account called “pbsdata”

  • wget https://github.com/openpbs/openpbs/archive/refs/heads/master.zip

  • unzip master.zip

  • cd openpbs-master/

  • apt install gcc make libtool libhwloc-dev libx11-dev libxt-dev libedit-dev libical-dev libncurses-dev perl postgresql-server-dev-all postgresql-contrib python3-dev tcl-dev tk-dev swig libexpat1-dev libssl-dev libxext-dev libxft-dev autoconf automake build-essential openssh-server net-tools

  • apt install expat libedit2 postgresql python3 postgresql-contrib sendmail-bin sudo tcl tk libical3 postgresql-server-dev-all

  • ./autogen.sh

  • ./configure --prefix=/opt/pbs

  • make ; make install

  • /opt/pbs/libexec/pbs_postinstall

  • edit /etc/pbs.conf and set PBS_START_MOM=1

  • chmod 4755 /opt/pbs/sbin/pbs_iff /opt/pbs/sbin/pbs_rcp

  • /etc/init.d/pbs start

  • . /etc/profile.d/pbs.sh

  • ps -ef | grep pbs_

  • qstat -Bf

  • pbsnodes -av ; pbsnodes -aSjv

Hope this helps

1 Like
3

Thank you, Adarsh I have installed it successfully. the procedure you mentioned is perfect. I will start working on my model. if I faced any problem kindly support give the suggestions form OpenPBS community team.

Thank you
Bill Nitzberg
OpenPBS Community Manager

1 Like
4

@adarsh Hi adarsh, with the above steps I was able to install head node.
How to go ahead with the compute nodes.

Using Ubuntu 20.04 . I tried with the execution deb file from other version - https://vcdn.altair.com/rl/OpenPBS/openpbs_22.05.11.ubuntu_20.04.zip

But after adding nodes,

it stays in unknown state. not able to figure it out.

5
  1. check with firewall / selinux disabled and system rebooted (if you disable selinux)
  2. DNS / hosts file is all good
  3. check pbs_mom service is up and running on the compute node ( ps -ef | grep pbs_ )
  4. check the mom logs
  5. check the server logs
  6. check ports 15001 to 15009 , 17001 are not blocked (between pbs server and pbs mom )
6

Hi Adarsh,

Just by adding info in /etc/hosts it was not working. I created a local DNS in head node and now PBS seems to be working fine.

Thanks a lot for your suggestions.

Regards,
Vinay

1 Like
7

Hi Adarsh,

I setup using your all blogs, can I know how to create multiple pbs job run uses

also I want to know how to added nodes to master head node

8

As root user on the headnode:

To add a node
qmgr -c “create node NODENAME”
#NODENAME is the hostname of the compute node

To delete a node
qmgr -c “delete node NODENAME”
#NODENAME is the hostname of the compute node

How to run jobs ? – As standard linux user
qsub -l select=1:ncpus=2:mem=1gb – /bin/sleep 1000 # to run on one node
qsub -l select=2:ncpus=2:mem=1gb -l place=scatter – /bin/sleep 1000 # to run on two nodes

9

Hi Adarsh
I am also following your all the steps, it is working for master node when I am adding one compute node in master showing state is down could you pls help this is my testing machine.

10

Please check

  1. pbs_mom process is up and running ( ps -ef | grep pbs_ )
  2. ports , firewall are not blocked
  3. selinux disabled and system rebooted
  4. hostname resolution (/etc/hosts)
11

Hi Adarsh,

Everything is working right now, I am able to submit the cpu based job but How can I configure and run gpu based test job or docker in Open PBS 23 version.
Pls guide for GPU based configuration. If possible pls ping your linkdin id.

12

Nice one !

For GPU configuration / scheduling please refer this

Please refer:

#######

Ref: 5.14.7.3 Basic GPU Scheduling