Unterschiede

Hier werden die Unterschiede zwischen zwei Versionen angezeigt.

--- hpc:tutorials:singularity:docker [2024/04/19 12:31] – gelöscht - Externe Bearbeitung (Unbekanntes Datum) 127.0.0.1
+++ hpc:tutorials:singularity:docker [2024/04/19 13:46] (aktuell) – jan.e
@@ Zeile 1: / Zeile 1: @@
+==== Using Docker Images ====
+== HOW THIS TUTORIAL WORKS ==
+This tutorial is written in order to give a small insight in using Docker image files with Singularity.
+It will cover, connecting to cluster, pulling the image and starting a database for a small test script.
+**It is written as shell script itself and might just be executed on the cluster!**
+You may also just copy paste the commands for a better understanding.
+Please check [[hpc:tutorials:singularity:create_image_files|this tutorial]] if you are interested in creating an own image file.
+Thx for reading \\
+Jan Eberhardt
+<code>
+#! /bin/bash
+# 0. Login on Frontend (you probably already did that)
+# Use your TUB account and host gateway.hpc.tu-berlin.de
+# ssh "<TUB account name>@gateway.hpc.tu-berlin.de"
+# 1. Get Docker Image
+# Go to your home directory and download the image via singularity.
+# You must load the singularity module beforehand.
+module load singularity/3.1.0
+# Pulling docker images is done by Singularity's pull command. Source will be something like "docker://[package name]".
+# Singularity will automatically download the latest version of the image and rewrite it to a Singularity image file (sif) as "[package name]_latest.sif".
+# Therefore you will need write permission in your current working directory (which is why we changed into home).
+cd
+singularity pull "docker://mongo"
+# 2. Create Python environment [if using Python]
+# Most Python projects will use open source libraries installed by pip. Since normal users are not allowed to do so, it is recommended to install
+# pip packages in user space or in a virtual python environment. We would discourage you from using user space for installation since most packages
+# you will only use once in your life and it is therefore cleaner to get an unique environment for each project of yours.
+# a) Load python module.
+module load python/3.7.1
+# b) Create the environment
+py="~/mongodb_venv"
+python3 -m venv ${py}
+# c) Install required pip packages and updates and create start script.
+# You may change the next steps accordingly to your project.
+# EXAMPLE PIP PACKAGE LIST
+#>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
+cat << EOL > "${py}/pip-packages"
+pymongo>=3.8.0
+EOL
+#<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
+# EXAMPLE PYTHON SCRIPT
+#>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
+cat << EOS > ~/mongodb_run.py
+#! python3
+from pymongo import MongoClient
+from pprint import pprint
+from sys import argv, executable
+from datetime import datetime
+print("Starting {:s}".format(argv[0]))
+print("Using environment {:s}".format(executable))
+print("Connecting to localhost")
+db_name = "test_database"
+col_name = "test_collection"
+client = MongoClient("localhost", 27017)
+print("Open collection '{:s}' on database '{:s}'".format(col_name, db_name))
+db = client[db_name]
+col = db[col_name]
+post = {
+"author": "HPC User",
+    "text": "This is a test record!",
+    "tags": [ "test", "mongodb", "pymongo" ],
+    "date": datetime.utcnow()
+}
+print("Inserting single record, resulting:")
+post_record = col.insert_one(post)
+pprint(post_record);
+EOS
+#<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
+"${py}/bin/pip3" install --upgrade pip
+"${py}/bin/pip3" install -r "${py}/pip-packages"
+# 3. Create DB directory
+dd="~/mongo"
+mkdir -p "${dd}"
+# 4. Start Server and run
+# Use mongodb_start.sbatch in order to allocate resources for and to start mongodb server:
+# This script will open up a server on a node and close it after the Python script finishes.
+#
+# We use the --exclusive switch of SBATCH in order to secure that port 27017 (default mongodb) is not in use.
+# If you do not like to use an exclusive node you will have to either accept the risk that the command fails or
+# to build a Singularity image of your own.
+#>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
+cat << EOF > ./mongodb_start.sbatch
+#!/bin/bash
+#
+# Start MongoDB docker image
+#
+#SBATCH --job-name=MongoDBStart
+#SBATCH --partition=standard
+#SBATCH --nodes=1
+#SBATCH --cpus-per-task=4
+#SBATCH --exclusive
+#
+#1 prepare
+module load singularity/3.1.0
+#- start instance (not the server)
+#- In that way we can use the instance command to stop the database when script finishes.
+singularity instance start --bind "${dd}:/data/db" ./mongo_latest.sif mongodb
+#- start server (by runscript)
+#- It will generate a lot of output, better redirecting that to oblivion (1>/dev/null).
+#- Also this call will lock your shell, avoided by ending the command with "&".
+singularity run instance://mongodb 1>/dev/null &
+#2 run program
+#- wait for database server to run
+sleep 5
+#- run script
+${py}/bin/python3 ~/mongodb_run.py
+#3 stop database after script finishes
+singularity instance stop mongodb
+EOF
+#<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
+sbatch ./mongodb_start.sbatch
+</code>