Skip to content

LRZ_Readme

Jump to bottom
kholzapfel edited this page Apr 24, 2023 · 10 revisions

LRZ Documentation

This project uses 2 different LRZ tools. The DSS is the storage backend, and the Compute Cloud hosts VMs. (Maybe also the Linux Cluster in the future.)

Overview

For STRAW and STRAWb, the STRAW-LRZ-VM runs 24/7, has 1 CPU with 4.5 GB RAM, and therefore synchronizes the module data from the ONC DB and monitors both detectors. But it can also do some simple calculations. For more power, another VM has to be set up. On the VM, the DSS-Storage is mounted to /dss.

DSS

Digital scientific storage (DSS) is s service provided by the LRZ (Munich). Here are the official docs for more information. You can access the DSS in two different ways.

  1. Mount the DSS container, e.g., on a VM hosted by the Compute Cloud. Important only a device with an mwn IP can mount the container. The section Mounting the DSS at a VM describes the required steps.
  2. Access the DSS via Globus which is a web-application similar to any cloud storage. The section Globus describes the required steps.

Management platform

At the DSS management webpage (only accessible within the TUM network, in case you use a VPN connection), persons with management permission see the DSS Containers. There you can manage the members, add NFS exports, and more.

Globus

A platform to share data like OwnCloud, NextCloud, DropBox, GoogleDrive, or similar, which is linked to the DSS. At the Globus webpage, you can log in with your LRZ credentials. By default, your account isn't granted access to the DSS. To do so, follow the registration section. Otherwise, you continue with access the DSS section

Registration

Before you can use Globus the first time, you must register that your LRZ credential gets the privileges for Globus. Login once to the DSS management webpage (only accessible within the TUM network, in case you use a VPN connection). The registration can take a while until it's transferred to the Globus server. Afterward, you can log in with your LRZ credentials and access the DSS at https://app.globus.org/

Old Registration Process

Previously, the Credential Registration Service for Globus data transfers must have been done manually. The webpage is offline, but this section holds the old process for reference.

Access the DSS

When your registration is successful, you can select the File Manager-tab (default after you log in at Globus):

Collection Leibniz Supercomputing Centre's DSS - CILogon
Path /dss/dssfs03/tumdss/pn69gu/pn69gu-dss-0004/

and set a bookmark for faster access the next time. (the path changed in history, so maybe check the dssweb (only available in the LRZ network <-> VPN))

Compute Cloud - VM hosting

The Compute Cloud is a VM hosting service of the LRZ. Here are the official docs. By default, a project has 10 CPU cores with 4.5 GB RAM per Core. The RAM ratio is fixed, but the total number of CPU cores can be higher. Therefore, contact the LRZ service desk.

The VMs among projects share the same hardware. Once a VM is active, it blocks the resource. Therefore, it is essential to only activate (un-shelf) the VMs if needed and deactivate (shelf) them afterward. However, VMs with one Core per project can run constantly. Therefore, those VMs can do regular jobs which do not require much computation power and can be an entry point to (un-)shelf other VMs which do more intense jobs. (How this works in an automated manner has to be figured out.)

This chapter contains the following sections:

Hosted VMs - Parameters and shh_config entries

The list of actual hosted VMs.

VM Name resources <user_name> <ip_address>
STRAW-LRZ-VM Active 24/7, 1 CPU, 4.5GB di46lez 138.246.233.224
Computing can be shelved if not needed, 4 CPU, 18GB di46lez 138.246.236.150

SSH config entries

SSH config entries to add to ~/.ssh/config:

Host straw-lrz-vm
    User            di46lez
    HostName        138.246.233.224
    IdentityFile    ~/.ssh/cloud.key
  
Host strawbcomputing
    User            di46lez
    HostName        138.246.236.150
    IdentityFile    ~/.ssh/cloud.key

replace ~/.ssh/cloud.key with the path to your ssh key.

VM management-portal

The VM management portal is https://cc.lrz.de. To log in there, you need a special LRZ-ID similar to your normal LRZ-ID but different. For now, this is: di46lez (password is posted in Slack, search for the ID).

IPs

You can assign a floating IP to a VM. This IP will never change unless you disassociate (unlink) it from the VM.

You can manage the floating IPs at the management-portal-> Project -> Network -> Floting IPs. To add a new IP, click Allocate IP To Project. Unfortunately, the mwn_pool is not exposed and, therefore, not accessible from the internet, which is a significant security issue. In contrast, the internt_pool is accessible and exposed to the internet.

Key Pairs

You can add key pairs with management-portal-> Project -> Compute -> Key Pairs. All key pairs are added to newly created VMs only. For existing VMs, you must manually add the keys to each machine. For more instructions about the ssh key, see the dedicated section.

Access a VM via SSH

Access to a VM is restricted to ssh with key pair only. First, create a new key pair. If you don't know how to search the internet for it. In the following, the key To add a key, someone with access to the VM has to add your public part (.pub) of the ssh key to the file ~/.ssh/authorized_keys on the VM. Afterward, you can log in with

ssh -i <your_key.key> <user_name>@<ip-address>

Replace <your_key.key> with the path to your .key file (not the .pub). For example, the <user_name> and <ip-address> are in the table of hosted VMs.

SSH config

You can set parameter pairs in the ssh-config file to make the ssh command easier. The file is located on your computer at ~/.ssh/config and add the following lines:

Host <name_of_the_vm>
    User            <user_name>
    HostName        <ip-address>
    IdentityFile    <your_key.key>

Replace <name_of_the_vm> with any name you want. Save the file. Now the ssh command reduces to:

ssh <name_of_the_vm>

In addition, the <name_of_the_vm> works with rsync and scp.

Some examples of using a VM

Copy files from/to the VM with rsync or scp

To sync files between different machines (but also on the same machine), you can use scp or rsync. The usage, especially with an ssh-config is straightforward, e.g., to download/upload files from/to the VM:

rsync <name_of_the_vm>:"/path/to/files/*.txt" target/dir  # downloads files all txt-files from /path/to/files/
rsync target/dir/*.txt <name_of_the_vm>:"/path/to/files"  # upload files all txt-files from target/dir/
rsync <name_of_the_vm>:"/path/to/files" target/dir/files  # download the whole directory
rsync <name_of_the_vm>:"/path/to/files/" target/dir  # equals the line above

For directories, you have to set the source string in the correct syntax. Without or with a tailing /, makes the difference. Compare the last two lines of the example. Furthermore, many documents - like - explain the difference and the various options.

Jupyter Notebook

A jupyter notebook server runs on the STRAW-LRZ-VM at port 8080. To use it on your computer:

  1. open a tunnel from your computer: ssh -L 8080:localhost:8080 <name_of_the_vm> and leave the terminal open, incl. the ssh connection.
  2. open http://localhost:8080 in the browser, PW: strawb

(Re-)start the server on the VM

Log in to the VM via ssh -L 8080:localhost:8080 straw-lrz-vm or ssh straw-lrz-vm; it does not matter. For the restart, ensure the old process isn't running at the target port. For this, get the PID with, e.g., ps -ef | grep jupyter-notebook or htop or similar. In case, kill it with kill <PID>.

jupyter notebook --no-browser --port=8080&; disown

From &; disown; & will put the task in the background, and disown unlinks it from the user, as closing the ssh connection will stop all linked processes.

Ports

You can also use any other port on your end or the VM, e.g., if you connect to different machines at once. On the VM, start the server with an updated --port=<VM_PORT> ; <VM_PORT> can be any unused port The tunnel:

ssh -L <LOCAL_PORT>:localhost:<VM_PORT> <name_of_the_vm>

has to match the <VM_PORT> and <LOCAL_PORT> can be any unused port + http://localhost:LOCAL_PORT

Clone this wiki locally