4. Logging in
Please make sure you have enabled your DUO device using UCI’s Duo infrastructure
To connect to an HPC3 login node use:
- server name:
hpc3.rcic.uci.edu
- login name:
your UCInetID
- password:
your password associated with your UCInetID
The following login methods are available, most common listed first:
4.1. Using ssh
SSH is the only way to directly log in to HPC3 for interactive use and we require multi-factor authentication for all password-based logins.
We describe two main methods below.
4.1.1. Method I: Password authentication
Password authentication with automated DUO push on your phone is the most common method for authentication. It requires your phone to have internet access to receive the push notification from DUO and to send your approval/denial back to DUO’s servers. The communications goes as :
You access HPC3 via your favorite SSH (SCP, SFTP) client from your laptop
HPC3 prompts you for a password and requests to use DUO authentication.
You then respond to the DUO app (DUO push) on your phone or your other DUO-enabled device.
Step by Step
Run ssh command
Your login name can be specified as ether user@hostname or given with the -l option, for example a user with UCInetID panteater can use one of the following:
ssh panteater@hpc3.rcic.uci.edu ssh hpc3.rcic.uci.edu -l panteater
When prompted for a password please enter password followed by Return key. Note, password will not be visible when typed:
Password:
Respond to multi-factor authentication prompts
Next, you will be prompted to enter a code (backup or generated by your DUO device) or request a push to your enrolled DUO-enabled device. A prompt looks similar to:
Duo two-factor login for panteater Enter a passcode or select one of the following options: 1. Duo Push to XXX-XXX-1234 Passcode or option (1-1):
Type desired option (in this example 1).
Passcode or option (1-1): 1
Use DUO on your phone
Now use the DUO app on your phone and respond to the received DUO notification. Press Approve on your DUO app when prompted. If the DUO authentication is successful you will see on your laptop:
Success. Logging you in... Last login: ....
After a successful login you will see a screen similar to the following:
+-----------------------------------------+ | _ _ _ _ ____ | | | | ___ __ _(_)_ __ (_) | ___| | | | |/ _ \ / _` | | '_ \ _____| | |___ \ | | | | (_) | (_| | | | | |_____| | |___) | | | |_|\___/ \__, |_|_| |_| |_|_|____/ | | |___/ | +-----------------------------------------+ Distro: Rocky 8.7 Green Obsidian Virtual: NO CPUs: 40 RAM: 191.8GB BUILT: 2022-08-30 14:02 ACCEPTABLE USE: https://rcic.uci.edu/documents/RCIC-Acceptable-Use-Policy.pdf [user@login-x:~]$
4.1.2. Method II: Key-based authentication
Attention
Every user-generated ssh key MUST have a non-empty passphrase. It is a requirement per our Acceptable use policy
NEVER add a different user’s ssh public into your authorized_keys file. This is a violation of account sharing.
Generate a different private key and password for each device you use for accessing HPC3. For example, if you two different laptops, generate a private key for each laptop.
Treat all of your ssh private keys with care. If you are on a shared system (e.g. a lab workstation), make sure that file permissions are set such you (and only you) can read and unlock the key with its passphrase.
If you don’t want to keep re-entering your passphrase, you should learn how to manage your ssh keys with the help of ssh agents. This provides a convenience of a passwordless ssh key, but has all the security of a password-protected key.
See Tutorials for SSH links.
Step by Step
Generate your ssh keys
This step is done once. Depending on your laptop, use the following guides:
- macOS:
- Linux:
- Windows (PuTTY):
- Windows (PowerShell):
The guides also explain how to copy your public SSH keys.
Once you generated your SSH keys and copied the public key to HPC3:
your laptop or workstation from which you are initiating ssh will have a passphrase protected ssh private key and a corresponding public key.
your public ssh key is added on HPC3 to your $HOME/.ssh/authorized_keys file.
Use ssh command to login
Once your keys are setup simply use
ssh
commands. For example a user with UCInetID panteater can use one of the following commands and provide the ssh passphrase when prompted:ssh panteater@hpc3.rcic.uci.edu ssh hpc3.rcic.uci.edu -l panteater
The above commands assume using default ssh keys which are usually id_rsa, id_dsa, id_ecdsa, id_ed25519, or identity files in your ~/.ssh/ directory.
If your ssh keys have non-default names and you do not have mapping in your ssh configuration file ~/.ssh/config, or if your ssh client does not use this configuration file (FileZilla), then you need to specify ssh key via -i flag:
ssh -i ~/.ssh/key-to-hpc3 panteater@hpc3.rcic.uci.edu ssh -i ~/.ssh/key-to-hpc3 hpc3.rcic.uci.edu -l panteater
4.1.3. Ssh and Xforward
If you want X-windows graphics to be forwarded through your ssh connection, then you should use the -X option in your ssh command, for example one of the following:
ssh -X panteater@hpc3.rcic.uci.edu
ssh -X hpc3.rcic.uci.edu -l panteater
Once logged in check if your environment variable DISPLAY is set:
[user@login-x:~]$ echo $DISPLAY
DISPLAY=localhost:27.0
If the Xforward is enabled in your ssh connection the DISPLAY will be set to a similar string, otherwise the output will be empty (no Xforward).
Note
Mac users must have XQuartz (standard application) installed in order to use XForwarding.
4.2. Using FileZilla and DUO
If you use FileZilla (Windows users) for files transfer you will need to change your login type to interactive in FileZilla Site Manager settings. For exact instructions please see Filezilla Site Manager

Fig. 4.1 Set login type
FileZilla with SSH keys
Sometimes it is more convenient to use SSH keys based authentication (in place of DUO). There are three mechanisms for use of the FileZilla client with SSH-2 keys.
Once you generated your ssh keys see the detailed instructions for the FileZilla SSH key based authentication mechanisms to setup your FileZilla client.
4.3. Using MobaXterm and DUO
Make sure that in your MobaXterm SSH tab -> Advanced ssh settings your Remote Environment is set to Interactive shell:
Fig. 4.2 Advancesd ssh settings
Do NOT enable remote server monitoring!
Remote server monitoring is an experimental feature of MobaXterm that runs unnecessary multiple processes on login node under your account.
These processes add to the overall load on the cluster while none of the information they collect can be used in any way for your work.
Check the lower portion of your MobaXterm window for the monitoring status, see the following figures for a reference:
Monitoring enabled - WRONG! The window shows info collected (outlined in red for clarity):
Monitoring disabled - correct!
To disable remote monitoring:
go to Settings > Configuration > SSH
under section SSH-browser settings ensure the Remote-monitoring (experimental) option is not selected.
4.4. Using VSCode
Attention
We do not allow running VSCode on login nodes because VSCode usage can result in login nodes becoming unusable by all. Any VSCode server instances will be removed from login nodes without a notice.
For users who desire to use VSCode RCIC supports the following method to run the VSCode server on compute nodes as a Slurm job and connect to it from their laptops. This is the only accepted method to run VSCode on HPC3.
VSCode remote server support requires ssh. To make things work smoothly, you must set up ssh key-based authentication from your laptop to HPC3.
There are two major parts to running the VSCode server on a compute node and connecting to it from your laptop:
- Part 1:
You need to submit a Slurm job specific to VSCode. This starts, on the assigned compute node, a user-specific sshd that is only available to that user.
- Part 2:
You need to configure your laptop VSCode client to communicate with this job’s sshd. Once set up properly, VSCode remote server development option performs all the work.
Please follow the instruction steps below to setup your VSCode connection on compute nodes.
Use
ssh
to connect to a cluster, see Method II: Key-based authentication to setup key-based authentication to HPC3. Critical: the ssh-key you setup must be protected with a password.Submit a batch job to set up a user-level sshd daemon on compute node which is needed for starting VSCode server.
[user@login-x:~]$ sbatch /opt/rcic/scripts/vscode-sshd.sh Submitted batch job 21877983
Slurm returns a job ID (in this example 21877983). Wait for the batch job to start running, the status in
squeue
output must be R:[user@login-x:~]$ squeue -j 21877983 JOBID PARTITION NAME USER ACCOUNT ST TIME CPUS NODE NODELIST(REASON) 21877983 standard vscode-s panteater panteater R 0:04 1 1 hpc3-22-09
Note
If you need additional resources, you can add the request when you run sbatch. For example, if you require 4 cpus:
sbatch --cpus-per-task=4 /opt/rcic/scripts/vscode-sshd.sh
Once the job starts running check its output file vscode-sshd-<jobID>.out in the directory where you submitted the job. There will be lines that look similar to:
Host hpc3-* HostName hpc3-22-09 Port 6666 ProxyJump panteater@hpc3.rcic.uci.edu User panteater UserKnownHostsFile /dev/null StrictHostKeyChecking no
Note, HostName will show a compute node name and the Port will show a port number. You will need to use them in the next steps.
This step needs to be done once and it will be used for all future invocations of VSCode on HPC3.
On your laptop in your $HOME there is a directory .ssh which was created when you enabled your ssh keys. Using a text editor, create a file .ssh/config with the following content:
Host hpc3-* HostName %h Port XXXX ProxyJump UCInetID@hpc3.rcic.uci.edu User UCInetID UserKnownHostsFile /dev/null StrictHostKeyChecking no
Replace UCInetID with yours, and XXXX with the Port number from vscode-sshd-<jobID>.out.Do not change any other other lines!If you already have .ssh/config file, simply add the content to it.
Note
Each time you start a new vscode-ssh.sh job, the Port number may change. This happens because the the vscode-sshd.sh looks for the first available network port within a pre-defined range on the specific compute node assigned to your job. Since that choice is dynamic, it can change for each new vscode-sshd.sh job. Simply edit your local .ssh/config to update the port to the port of your currently-running vscode job that you started in Step 3.
On your laptop start your VSCode application.
Note, images below show VSCode application for macOS, the Windows version may look slightly different but the concept is the same.
5.1 Click on the open remote window icon and choose Connect to Host… from the menu
Fig. 4.3 Connect to host
5.2 Choose + Add new SSH Host… from the menu:
Fig. 4.4 Add new ssh host
5.3 In the Enter SSH Connection Command box, enter the compute node name fromthe output file of your submitted batch job, then press Enter or Escape key:Fig. 4.5 Enter SSH Connection command
5.4 In the Select SSH configuration file to update menu choose yourlocal .ssh/config (use local file from your user area on your laptop):Fig. 4.6 Enter SSH configuration file to update
5.5 When the window updates press Connect button:
Fig. 4.7 Press connect
5.6 In a new window you will be asked to provide your ssh passphrase(two times), type it where indicated by your Application:Fig. 4.8 Provide ssh credentials
Once the authentication is successful you will see the changes on the lower portion of the window, they indicate that the connection is getting established and the server is getting setup (shown with blue outline above). It may take a few minutes for the VSCode to setup the server.
5.7 Once done, you will see the open remote window icon showing compute node name (in blueoutline). This means your connection is ready and you can proceed with your work as usual:Fig. 4.9 Connection is ready
Shutting down your remote VSCode server
Attention
The remote start of VSCode leaves the server running long after you have finished your work and closed your remote connection. You need to shutdown your server properly.
Shutting down your remote server is a simple 2-step process:
- Step 1 on your laptop:
in VSCode application choose File > Close Remote Connection and follow your application prompts to disconnect from the host.
- Step 2 on login node:
cancel your VSCode job (by your jobID):
[user@login-x:~]$ scancel 23383635
If you don’t cancel your job it will continue consuming your lab or your personal Slurm allocation balance.
Reconnecting to an already running VSCode Server
If you have not shutdown the server in Step 6 above, you simply re-open the connection as you did in Step 5. Please remember, you Slurm job continues to charge your account as long as it consumes resources.