ICS Instructional Openlab Linux Cluster

Summary

The ICS Instruction Openlab Linux Server cluster is for general purpose computing at the School of ICS and accessible to any user with an ICS shell account. Cluster priority is given to instruction, then undergraduate students, then graduate students, and then ICS researchers.

Guidelines

In order to ensure that Openlab cluster remain available to instruction and researchers fairly, users may expect the following:

  • Long running jobs must be submitted through our SLURM job queueing system in order to balance resources.
  • Jobs executed from via SLURM queueing system will be given priority over jobs running on the command line.
    • Processes running for longer than 2 hours will be reniced 19.
  • Processes running for longer than 5 days will be suspended.
  • Processes for users with more than 100 processes will be suspended.
  • Files in the /tmp directory will be cleared after two weeks.

Please use slurm for long running or serial projects requiring more than one openlab node. Processes running through slurm are not subject to the above guidelines.

Instructors may reach out to heldpesk@ics.uci.edu in order to request course exceptions to any of these guidelines.

Also see Linux Openlab FAQ

Home Directory

Your ICS home directory is shared across all ICS support linux hosts including the openlab hosts.

Connecting to Openlab

You may log in remotely from any host using an ssh client. See our common tools page for a list of clients.

Login into one of the following domain names (e.g. ssh openlab22.ics.uci.edu). In this case though, openlab and openlab22 will take you to the same set of servers since we have moved the same OS for all openlab machines.

Domain Name Operating System
openlab.ics.uci.edu Ubuntu 22.04 hosts only
opengpu.ics.uci.edu Ubuntu 22.04 hosts only

NOTE: If you are running a long running process, please take note of the hostname you are running on. The name “openlab.ics.uci.edu” is a load balancer and may not point you to the same host twice. You can determine the hostname by running the command `hostname` from the shell cli. When you come back to check on your process, you will ssh into this name instead of openlab.ics.uci.edu.

MOD/Splash Screen

When you login to the Openlab Linux cluster you will receive a lot of information that includes sections about fair use, system load and message from individual software ICS softward modules you add to your environment. This information may also include a message similar to the following:

Expanded Security Maintenance for Applications is not enabled.
  
19 updates can be applied immediately.
11 of these updates are standard security updates.
To see these additional updates run: apt list --upgradable

24 additional security updates can be applied with ESM Apps.
Learn more about enabling ESM Apps service at https://ubuntu.com/esm

You may safely ignore this message. This message is automatically managed by Ubuntu Landscape and regards the host that you have logged into, not your local system. This message indicates that there are patches available or that patches that have been applied require a system restart in order to fully install. Our systems administrators will take care of that, typically between quarters when systems are not heavily used.

Off Campus Access

ADVISORY: Off campus access requires ssh keys. Users connecting from off campus must use keys. Users connecting from off campus cannot use passwords to login to the openlab cluster.

What does ICS consider on campus:

  • Any host with a reverse dns name that ends in uci.edu. This includes most campus hosts and hosts connected via VPN.
  • Any host connected to the campus 172.23.0.0/20 network.

These are the match rules in the ICS sshd_config file:

Match Host *.uci.edu
  PasswordAuthentication yes
Match Address 172.23.0.0/20
  PasswordAuthentication yes

Load Balancing

The name openlab.ics.uci.edu takes you to a load balancer. The load balancer will send your connection request to the machine with the lowest load.

If your work requires you to work with multiple shells on the same host, then login directly using a hostname instead of openlab.ics.uci.edu

Determining Hostname

In order to identify which openlab host you are on run the command:

hostname

In the following example, the user has logged into a host named circinus-30.ics.uci.edu:

% hostname
circinus-30
SSH Tunneling

In order to create an ssh tunnel to the host that you are a process on, you cannot use the name openlab.ics.uci.edu.

In the following example, the user has connected to openlab and run jupyterlab on port 5555. They have run hostname and determined that the name of the host they have run jupyterlab on is circinus-2.ics.uci.edu.

% ssh -L 8080:localhost:5555 luser@circinsus-2.ics.uci.edu

Availability

The openlab cluster is available on and off campus via SSH and SLURM.

There are limitations regarding authentication and how users can connect to the openlab cluster. Please refer to the following matrix for specific information:

LocationSSH with KeysSSH with PasswordSLURM
On campus Yes Yes Yes
Off Campus* Yes No No
UCI VPN Yes Yes Yes

*Off-campus locations include student housing, University Hills housing, and the campus WIFI network.

Using UCI VPN

Check here for directions connecting to the UCI Campus VPN. Make sure you download the software version and select “UCI” from the dropdown menu when choosing a connection type.

Windows Users

  1. Newer versions of Windows have ssh built in. Optionally, you can install Linux Subsystem for Windows to take advantage of running other Linux commands locally.
  2. Open terminal, using the updated Windows Terminal is recommended.
  3. ssh {ics_userid}@openlab.ics.uci.edu

or

Please reference this page on how to connect to Openlab using Putty software: set up Putty for Openlab.

Mac Users

Open Terminal and type:

ssh yourICSusername@openlab.ics.uci.edu

FOSS SSH Clients

You may use any SSH client to connect to the openlab servers. This short list provides a FOSS ssh clients:

See our common tools page for a list of ssh clients.

Setup your SSH Keys

Off-campus users must use SSH keys to connect to the openlab cluster. Follow these directions to Set Up SSH Keys (includes a link to PuTTY key setup).

Donations

Much of the openlab cluster runs on donated hardware. Feel free to contact support if you think you have some spare hardware that you would like to add.

Details

  • Operating System: Ubuntu 22.04.
  • Provisioning: Puppet 6.x
  • Primary Investigator: Computing Support
  • Contact Email: helpdesk@ics.uci.edu

Hardware Specifications

The Openlab hosts are made up of two sets of hardware that are grouped with the following machines:

Circinus Cluster

Description Information
ClusterCircinus Cluster
Count 96
Model HP Proliant SL390 G7
Processor 2 x Intel Xeon X5680 @ 3.33 Ghz
RAM 96 GB
HDD Varies
RAID No
Redundant Power Yes

Odin

Description Information
ClusterOpenlab Cluster
Count 1
Model Dell PowerEdge R815
Processor AMD Opteron™ Processor 6378
RAM 512 GB
HDD 98GB
RAID 5
Redundant Power Yes

Tristram

Description Information
ClusterOpenlab Cluster
Count 1
Model Dell PowerEdge R815
Processor AMD Opteron™ Processor 6378
RAM 512 GB
HDD 98GB
RAID 5
Redundant Power Yes

User Access

  • All Users

Services Provided

Code Editor suggestions

Members

Complete rosters for the openlab cluster can be displayed with the following sinfo command:

module load slurm
sinfo

Long Running Processes

Resuming Suspended Processes

If you find your process has been suspended, you may resume it by using the kill command ton send the CONT signal.

kill -CONT <pid>

You may list your processes using the command `ps ux`. Suspended processes will have a “T” in the 8th column

USER         PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
hans     2458438  0.0  0.0   6476  2244 pts/18   S+   14:08   0:00 grep --color=auto --exclude-dir=.bzr --exclude-dir=CVS --exclude-dir=.git --exclude-dir=.hg --exclude-dir=.svn --exclude-dir=.idea --exclude-dir=.tox T
hank     3554779  0.0  0.0  20480  3164 ?        Ts    2022  24:43 SCREEN -xxR

In this case you would resume the SCREEN process, pid 3554779 by running:

kill -CONT 3554779

Exceptions

The following long running processes will not be automatically suspended on the Openlab Linux Cluster:

  • bash
  • rbash
  • csh
  • dash
  • ksh
  • ksh93
  • tcsh
  • zsh
  • screen
  • tmux
  • nologin
  • nologin
  • ksh
  • rksh
  • csh
  • (sd-pam)
  • sshd
  • ssh-agent
  • systemd
  • dbus-daemon
  • node
  • vim
  • remote-dev-serv
  • fsnotifier
  • ion.clangd.main
  • ld-linux-x86-64
  • fsnotifier
  • cpptools
  • cpptools-srv
  • xdg-mime
  • dbus-send

Troubleshooting

RSA host key for <hostname> has changed

If you receive an error “Hot key verification failed” when logging in, you may, if you are confident you are logging into the correct host, clear the offending key by issuing the following command, on most platforms, to clear the old host key and resolve the error:

ssh-keygen -R <hostname>

If this does not work on your platform, please send a message to helpdesk@ics.uci.edu along with the output from the command so we can extend this information.

RC Files

ICS users may occasionally damage their dot files, which results in unusual-looking SSH prompts and/or being unable to use the “module” command.

If this is the case, click here for directions to https://swiki.ics.uci.edu/doku.php/accounts:restore_unix_dot_files If the above fails, contact helpdesk.

hardware/cluster/openlab.txt · Last modified: 2024/04/19 10:34 by hans
CC Attribution-Noncommercial-Share Alike 4.0 International
Driven by DokuWiki Recent changes RSS feed Valid CSS Valid XHTML 1.0