Installation

This page describes the installation procedure of the Web Interface and the GDS back-end.

Both, the back-end and the web interface, are integrated into the SeisComP environment. The following tables shows the location of the GDS bundle:

Location

Description

$SEISCOMP_ROOT/bin/gds

Binary of the back-end server

$SEISCOMP_ROOT/bin/quakelink

Binary of the QuakeLink server

$SEISCOMP_ROOT/share/gds/tools

Python framework for content filter and spooler scripts as well as sample implementations

$SEISCOMP_ROOT/share/gds/web

Base directory of the Web Interface

Supported Operating Systems

This installation manual was tested under the following operating systems.

Distribution

Versions

RHEL

8 / 9

Ubuntu

22.04 / 24.04

Debian

11 / 12

Distributions not listed above may work but are not officially supported. If you setup GDS from scratch consider to upgrade to the latest supported OS version first.

Dependencies

The GDS relies on the following software packages:

Dependency

Required by

Python >= 3.10

web interface

Python >= 3.6

back-end

NGINX web server including Gunicorn

web interface (optional)

QuakeLink Server

back-end, web interface

DBMS (see table below)

back-end, web interface

In addition the following Python libraries are required by the web interface:

  • Django = 5.2.*

  • Django-Bitfield

  • Django-Extensions (optional)

  • Django-Import-Export (optional) - Optional Tablib data formats: ods,xls,xlsx,yaml

  • Django-SimpleHistory

  • MariaDB or PostgreSQL connector

  • Gunicorn (optional)

Finally one of the following Database Management Systems (DBMS) is required:

DBMS

Minimum Version

Comment

MariaDB

10.4

PostgreSQL

12

MySQL

8

Not covered by this manual but similar to MariaDB

SQLite

3.21

Not covered by this manual but example configuration in $SEISCOMP_ROOT/share/gds/web/gds/settings.py

The installation of a web server is optional since Django ships with a tiny build-in web server. This build-in server is sufficient for testing purposes and if the GDS should be configured respectively monitored only locally (same machine). To use the full potential of the web configuration approach, the integration into an NGINX web server is recommended.

Note

In addition to the dependencies listed above more packages are required by the SeisComP environment. Please refer to the SeisComP manual. Since SeisComP version Seattle dependencies may be installed via seiscomp install-deps. Issue seiscomp help install-deps for a list of available packages. Furthermore you might want to use the same database management system for the GDS database as used by the SeisComP messaging system to safe computational resources.

Python

System Packages

Installation of Python, the package installer for Python (PIP) and the virtual environment module depending on your operating system:

# Ubuntu / Debian 12
sudo apt install python3 python3-pip python3-venv

# RHEL
su -
dnf install python3.11 python3.11-pip

# Debian 11
# Install Python 3.11 from source (not covered here)

Virtual Environment

A virtual environment is created which will host all dependencies of the GDS front-end.

mkdir -p $SEISCOMP_ROOT/var/venv/gds
cd $SEISCOMP_ROOT/var/venv/gds

# Ubuntu/Debian
python3 -m venv web
web/bin/pip install --upgrade pip

# RHEL
python3.11 -m venv web
web/bin/pip install --upgrade pip

NGINX

The NGINX setup is optional if you are not planning to expose the GDS web-interface to the public. Nevertheless an installation even for internal use recommended to serve all web modules though one port.

# Ubuntu/Debian
sudo apt install nginx

# RHEL
su -
dnf install nginx

Database

Install one of the following database management systems (DBMS):

For production environments it is also recommended to install the NGINX web server.

MariaDB

Note

The default MariaDB version under RHEL 8 is 10.3. If you already run MariaDB in version 10.3 you need to upgrade to version 10.5

# Ubuntu/Debian
sudo apt install mariadb-server mariadb-client libmariadb-dev pkg-config python3-dev gcc
$SEISCOMP_ROOT/var/venv/gds/web/bin/pip install mysqlclient

# RHEL 9
su -
dnf install mariadb-server pkg-config python3.11-devel gcc
dnf install mariadb-devel --enablerepo=devel
exit
$SEISCOMP_ROOT/var/venv/gds/web/bin/pip install mysqlclient

# RHEL 8
su -
dnf module reset mariadb
dnf module enable mariadb:10.5
dnf distro-sync
dnf install mariadb-server mariadb-devel pkg-config python3.11-devel gcc
exit
$SEISCOMP_ROOT/var/venv/gds/web/bin/pip install mysqlclient

PostgreSQL

Note

The default PostgreSQL version under RHEL 8 is 10. If you already run PostgreSQL in version 10 you need to upgrade to version 12

# Ubuntu/Debian
sudo apt install postgresql postgresql-client
$SEISCOMP_ROOT/var/venv/gds/web/bin/pip install psycopg2-binary

# RHEL 9
su -
dnf install postgresql-server
exit
$SEISCOMP_ROOT/var/venv/gds/web/bin/pip install psycopg2-binary

# RHEL 8
su -
dnf module reset postgresql
dnf module enable postgresql:12
dnf distro-sync
dnf install postgresql-server
exit
$SEISCOMP_ROOT/var/venv/gds/web/bin/pip install psycopg2-binary

Django

The Web Interface requires the Django framework in version 5.2. as well as the following extensions:

  • django-bitfield

  • django-extensions

  • django-import-export

  • django-simple-history

# Ubuntu/Debian and RHEL
$SEISCOMP_ROOT/var/venv/gds/web/bin/pip install -r $SEISCOMP_ROOT/share/gds/web/requirements.txt

Web Interface

Database Initialization

The GDS database must be created and permissions have to be granted to a user.

MariaDB

Enable and start the database server.

sudo systemctl enable mariadb
sudo systemctl start mariadb

Secure the database installation (optional). Use default answers except when ask for the new root password.

sudo mysql_secure_installation

Create the database and grant access.

mysql -h localhost -u root -p
mysql> CREATE DATABASE gds CHARACTER SET utf8mb4;
mysql> GRANT USAGE ON gds.* TO sysop@localhost IDENTIFIED BY 'sysop';
mysql> GRANT ALL PRIVILEGES ON gds.* TO sysop@localhost;

If the MariaDB server resides on a different machine, the localhost must be replaced by the server’s IP. Also the database name, user and password may need to be modified.

Django requires MariaDB timezone information. Issue the following command to check if your MySQL database is timezone aware:

mysql> SELECT CONVERT_TZ(now(), 'UTC','Europe/Berlin');

If the result is NULL you need to populate your DBMS with timezone information:

mysql_tzinfo_to_sql /usr/share/zoneinfo | mysql -D mysql -u root -p
mysql -u root -p -e "FLUSH TABLES;" mysql

The GDS offers to log the full bulletin content in the database. Depending on the messages you plan to send and on the operating system you use the default maximum allowed database packet size might be to small. Use the following commands to review the current settings of your DBMS:

mysql -h localhost -u root -p
mysql> SHOW VARIABLES LIKE 'max_allowed_packet';

If you think you will hit this limit you need to increase the packet size value in your DB server configuration. E.g., for RHEL and MariaDB this is done in the /etc/my.cnf.d/server.cnf:

[mariadb]
max_allowed_packet = 16777216

PostgreSQL

On RHEL the database cluster needs to be initialized and the system service enabled and started first.

# RHEL
su -
postgresql-setup --initdb
/var/lib/pgsql/data/pg_hba.conf

host    all             all             127.0.0.1/32            md5
host    all             all             ::1/128                 md5

systemctl enable postgresql
systemctl start postgresql

Create a database user and a new database.

sudo -u postgres -i
createuser -P -d sysop
   Enter password for new role:
   Enter it again:
   Shall the new role be a superuser? (y/n) n
   Shall the new role be allowed to create more new roles? (y/n) n
createdb -O sysop gds
exit

Create and Adjust Settings

The web interface ships with 2 example configuration files. Change to the front-end directory and create a copy of those files.

cd $SEISCOMP_ROOT/share/gds/web
cp examples/settings*.py gds/

The $SEISCOMP_ROOT/share/gds/web/gds/settings.py holds the main configuration settings. Although the defaults in place will work for most installations it is a good practice to review them. Especially the connection settings for the database as well as the GDS and QuakeLink server should be checked.

The $SEISCOMP_ROOT/share/gds/web/gds/settings-production.py extends the default settings by options needed for a production environment using a reverse proxy, see Web Server Integration. Typically one may change the value of

  • FORCE_SCRIPT_NAME if you a are deploying the GDS on a sub-path other than /gds.

  • ALLOWED_HOSTS which must include all IPs and hostnames you will use to access the GDS web interface.

Note

The Django build-in web server is considered to be insecure and should not be exposed to the public. You may use it in production if your network is properly shielded.

Completing the Installation

The database must be initialized with the Django admin tables and the GDS model. Also a Django super user must be created to login to the web interface. The 3rd command (loaddata) is optional. It will create an example configuration containing an email service, a queue matching for all world-wide events and finally a subscription to this queue. The 4th command (collectstatic) is only needed for the production environment. It copies all static files into a single directory (STATIC_ROOT) which needs to be served by NGINX. The last command (generate_secret_key) generates a secret key used in production.

source $SEISCOMP_ROOT/var/venv/gds/web/bin/activate
cd $SEISCOMP_ROOT/share/gds/web
python manage.py migrate
python manage.py createsuperuser
python manage.py loaddata sample-data.json
python manage.py generate_secret_key > gds/secretkey.txt
deactivate

To test the configuration the Django build-in web server may be started:

cd $SEISCOMP_ROOT/share/gds/web
$SEISCOMP_ROOT/var/venv/gds/web/bin/python manage.py runserver

By default the server is available under http://localhost:8000

To run the Django build-in server with production settings and to serve on all network interfaces issue the following command:

cd $SEISCOMP_ROOT/share/gds/web
$SEISCOMP_ROOT/var/venv/gds/web/bin/python manage.py runserver --settings=gds.settings-production 0.0.0.0:8000

Note

The Django build-in web server is considered to be insecure and should not be exposed to the public. You may use it in production if your network is properly shielded.

Web Server Integration

If you verified that the build-in web server is running, you may stop it now (Ctrl+C) and continue with web server integration. We recommend a combination of NGINX and Gunicorn for this purpose.

NGINX

NGINX will be configured as a reverse proxy and relies on an external WSGI service, e.g., Gunicorn, to map HTTP requests to the Django web application. The communication between NGINX and Gunicorn will be realized through a Unix socket, although TCP based deployments are possible, e.g, if both processes should run on different machines.

  • Add the GDS upstream and location blocks to your NGINX configuration:

    • Debian/Ubuntu:

      • Edit /etc/nginx/sites-available/default.conf

      • Put the content of $SEISCOMP_ROOT/share/gds/web/examples/nginx/gunicorn/upstream.conf before the server block

      • Put the content of $SEISCOMP_ROOT/share/gds/web/examples/nginx/gunicorn/location.conf inside the server block

        upstream gds_app_server { ... }
server {
     ...
     location /gds/ { ... }
     location /gds/static { ... }
}

    • RHEL:

      • Copy the example files to the conf.d and default.d directories

        cd $SEISCOMP_ROOT/share/gds/web/examples/nginx/gunicorn/
su
cp upstream.conf /etc/nginx/conf.d/gds.conf
cp location.conf /etc/nginx/default.d/gds.conf

    Note

    If you are deploying the GDS on a path other than /gds, adjust the location directives accordingly. Make sure the sub-path matches the FORCE_SCRIPT_NAME variable in your $SEISCOMP_ROOT/share/gds/web/gds/settings-production.py.

  • NGINX will serve static files from STATIC_ROOT typically configured to /usr/share/nginx/html/gds/static in $SEISCOMP_ROOT/share/gds/web/gds/settings.py.

    Create the static directory, change ownership to the Django user and collect the files:

    sudo mkdir -p /usr/share/nginx/html/gds/static
sudo chown sysop:sysop /usr/share/nginx/html/gds/static
cd $SEISCOMP_ROOT/share/gds/web
$SEISCOMP_ROOT/var/venv/gds/web/bin/python manage.py collectstatic

  • If your server is not using SSL, which is not recommended, you have to change the following variables in $SEISCOMP_ROOT/share/gds/web/gds/settings-production.py

    # NOT recommended, configure NGINX with SSL instead
CSRF_COOKIE_SECURE = False
SESSION_COOKIE_SECURE = False

  • Test the NGINX configuration:

    sudo nginx -t

  • Enable and start NGINX:

    sudo systemctl enable nginx
sudo systemctl start nginx

Gunicorn

NGINX relies on an external WSGI service, e.g., Gunicorn, to map HTTP requests to the Django web application. The communication between NGINX and Gunicorn will be realized through a unix socket, although TCP based deployments are possible, e.g, if both processes should run on different machines.

Gunicorn will run under the same user under which Django is installed. The configuration examples use the default SeisComP user sysop for this purpose.

  • Copy the Gunicorn configuration file to the Django base directory:

    cd $SEISCOMP_ROOT/share/gds/web/
cp examples/gunicorn/gunicorn.conf.py .

    Adjust the configuration, e.g., logging level, path and format if necessary.

  • Copy the Gunicorn socket and service systemd configuration:

    sudo cp $SEISCOMP_ROOT/share/gds/web/examples/gunicorn/gunicorn-gds.* /etc/systemd/system

    Note

    If you use a user other than sysop or a home directory other than /home/sysop ajust the proroperties user, group, WorkingDirectory and ExecStart in /etc/systemd/system/gunicorn-gds.service accordingly.

  • Enable and start the service:

    sudo systemctl enable gunicorn-gds
sudo systemctl start gunicorn-gds

Trouble Shooting

  • Check if the socket file /var/run/gunicorn-gds.sock was created and ensure that the NGINX user (nginx or www-data) has write access to this file

  • Make a request directly at the socket:

    sudo -u nginx curl --unix-socket /var/run/gunicorn-gds.sock -H "SCRIPT_NAME: /gds" http

  • By default Gunicorn logs to $HOME/.seiscomp/log/gds-web/access,error.log

  • Reload gunicorn-gds service after modifications to your $SEISCOMP_ROOT/share/gds/web/gunicorn.conf.py or Django configuration file: sudo systemctl reload gunicorn-gds

  • Reload systemd configuration after modifications to systemd service and socket files: sudo systemctl daemon-reload

  • Reload the NGINX server configuration: sudo nginx -t && sudo systemctl reload nginx

Next Steps

  • Login using the administrator account you created

  • Create staff users with some/all privileges prefix by gds_

  • Create Services, Queues and Subscriptions

  • Configure (re)start the Back-end

Back-end

The GDS back-end server is integrated into the SeisComP environment and is started like every other SeisComP application. Before running the server, the database must be created and initialized. Also a $SEISCOMP_ROOT/etc/gds.cfg must be created with a proper Database Configuration and Service Configuration.

Database Configuration

Make sure that the correct database plugin is loaded. This can be done in $SEISCOMP_ROOT/etc/global.cfg or in SEISCOMP_ROOT/etc/gds.cfg if different from the DBMS used by SeisComP. Next the database URL must be specified as demonstrated in the following database specific sections.

MariaDB/MySQL

# List of plugins loaded at startup. Separate multiple names by comma. Add
# ${plugins} to consider all previously read values.
plugins = ${plugins}, dbmysql

# Define the database connection. If no database is configured (which is the
# default) and a messaging connection is available, the application will
# receive the parameters after the connection is established. Override these
# values only if you know what you are doing.
# The format is: service://user:pwd@host/database. "service" is the name of the
# database driver which can be queried with "--db-driver-list". IMPORTANT: This
# parameter replaces former "database.type" and "database.parameters" which
# have been removed.
database = mysql://sysop:sysop@localhost/gds

PostgreSQL

Note

If the postgresql_psycopg2 Django back-end is used an additional dbSequencePrefix must be configured.

# List of plugins loaded at startup. Separate multiple names by comma. Add
# ${plugins} to consider all previously read values.
plugins = ${plugins}, dbpostgresql

# Define the database connection. If no database is configured (which is the
# default) and a messaging connection is available, the application will
# receive the parameters after the connection is established. Override these
# values only if you know what you are doing.
# The format is: service://user:pwd@host/database. "service" is the name of the
# database driver which can be queried with "--db-driver-list". IMPORTANT: This
# parameter replaces former "database.type" and "database.parameters" which
# have been removed.
database = postgresql://sysop:sysop@localhost/gds

# By default a database sequence is named after the table, suffixed by "_seq":
# "[tablename]_seq". Sometimes the column name is used as well, e.g. when using
# postgresql_psycopg2 as Django database back-end. If set, the value of this
# parameter will be added to the sequence suffix:
# "[tablename][dbSequencePrefix]_seq"
dbSequencePrefix = _id

Service Configuration

For each Service, External Criterion and Filter a corresponding entry in the servers configuration file must exist. If you loaded the sample-data.json as described in ref:sec-inst-web-final then you may use the following configuration:

# Service directory (mandatory). For each service a folder must exist where
# ID.content and ID.address files are stored.
service.email.directory = @ROOTDIR@/var/lib/gds/email

# The full path to the filter application.
# Environment variables ('$NAME') and SeisComP path variables ('@NAME@') are
# resolved at time of application start. In addition the following event
# context variables ('@NAME@') are resolved at each command run: MATCH,
# REVISION, EVENTID, ORIGINTIME, LATITUDE, LONGITUDE, DEPTH, MAGNITUDE, RMS,
# PHASES, MAGNITUDECOUNT, EVALUATIONMODE, EVALUATIONSTATUS, REGION, EVENTTYPE
# and MAGTYPE.
# It is a good practice to use quotes for string variables as they may be empty
# or contain white spaces.
service.email.filter.primary.cmd = "@DATADIR@/gds/tools/filter_email.py -t 8"

# MIME type of the filter output. This parameter is used by the web front-end
# for content display.
service.email.filter.primary.mimetype = text/gds

# Timeout in seconds after which the command is aborted. Use a negative value
# to indicate an infinite timeout.
service.email.filter.primary.timeout = 10.0

# Spool command to execute.
service.email.spooler.cmd = "@DATADIR@/gds/tools/send_email.py -t 8"

# Polling interval in seconds the spooler will scan the service directory for
# new files.
service.email.spooler.interval = 2.0

# Timeout in seconds after which the command is aborted. Use a negative value
# to indicate an infinite timeout.
service.email.spooler.timeout = 10

If you followed front-end installation steps and loaded the sample-data.json then you might copy an example backend configuration to get you started:

cd $SEISCOMP_ROOT
cp share/gds/web/examples/gds.cfg etc/

Start the server with debug output and check for any problems:

seiscomp exec gds --debug

If no errors are found stop the server (Ctrl+C), then enable and start the SeisComP service:

seiscomp enable gds
seiscomp start gds

Update Instructions

If you already operate a GDS version and you intent to install an update you should always follow the general update instructions. Additional update steps may be required depending on the version you are upgrading from.

General Instructions

  • Read the CHANGELOG file to get an overview of the changes

  • Consider creating a database backup

  • Install the new GDS version via [GSM](https://docs.gempa.de/gsm/README) or extract the tar file overriding existing files in the installation folder.

  • Check for new or updated parameters in the $SEISCOMP_ROOT/share/gds/web/examples/settings.py and $SEISCOMP_ROOT/share/gds/web/examples/settings-production.py configuration files and apply them to your copies

  • Apply database migrations if any

  • Remove and collect static files again

  • Restart the GDS server

  • Restart Gunicorn

Database Migrations

The data model might have changed. Check for outstanding migrations and apply them if necessary:

cd $SEISCOMP_ROOT/share/gds/web
$SEISCOMP_ROOT/var/venv/gds/web/bin/python manage.py showmigrations
$SEISCOMP_ROOT/var/venv/gds/web/bin/python manage.py migrate

Static Files

The static files changed and need to be collected again:

rm -rf /usr/share/nginx/html/gds/static/*
cd $SEISCOMP_ROOT/share/gds/web
$SEISCOMP_ROOT/var/venv/gds/web/bin/python manage.py collectstatic

Migration to 2025.202 (Django 5.2, venv)

With version 2025.202 the GDS front-end switched from Django 3.2 to version 5.2 (LTS). As a consequence a minimum Python version of 3.9 is required and the minimum database versions were increased. Affected operating systems are:

  • RHEL 8 (EOL 2029)

    • Upgrade to Python 3.9

    • Upgrade to MariaDB 10.5 (available with RHEL 8.4)

    • Upgrade to PostgreSQL 12

  • Ubuntu 20.04 (EOL 2025, not longer supported by this manual)

    • Upgrade to Python 3.9

    • Upgrade to MariaDB >= 10.4

In addtion the recommended setup is now based on a Python Virtual Environment (venv). If your OS is Debian, Ubuntu >=22.04 or RHEL 9 then you may continue without a venv and just update the Python dependencies in addition to the general update instructions. Otherwise you will need to switch to Python 3.9 using a venv.

Migration to 2021.294 (Django 3.2, simple-history)

With version 2021.294 the GDS front-end switched from Django 3.1 to the LTS version 3.2. Also history support for the data model has been added allowing to review all changes to a particular configuration object and to revert back to a previous revision.

Python Dependencies

Update Django and install the SimpleHistory django plugin:

python3 -m pip install --user django==3.2.*
python3 -m pip install --user django-simple-history==3.0.*

Update of Configuration Files

Compare your $SEISCOMP_ROOT/share/gds/web/gds/settings.py with the updated $SEISCOMP_ROOT/share/gds/web/gds/example-settings.py. Especially make sure to

  • Add the line DEFAULT_AUTO_FIELD = 'django.db.models.AutoField'

  • Add 'simple_history', to the INSTALLED_APPS list

  • Add 'simple_history.middleware.HistoryRequestMiddleware', to the MIDDLEWARE list

Compare your $SEISCOMP_ROOT/share/gds/web/gds/settings-production.py with the updated $SEISCOMP_ROOT/share/gds/web/gds/example-settings-production.py. Especially make sure to

  • Add 'simple_history', to the INSTALLED_APPS list

Database Migrations

The data model changed. To complete the update run:

cd $SEISCOMP_ROOT/share/gds/web
python3 manage.py migrate
python3 manage.py populate_history --auto

Static Files

The static files changed and need to be collected again:

cd $SEISCOMP_ROOT/share/gds/web
rm -rf static
python3 manage.py collectstatic

Migration to 2020.282 (Django 3.1)

With version 2020.282 the GDS front-end switched from Django 1.11 to 3.1 and requires a minimum Python version of 3.6. The back-end was updated to support Python 3 but continues to support Python 2 starting from version 2.7 in order to support SeisComP3 Jakarta 2018.327 and older.

Cleanup

Since this is a major update it is recommended to move your $SEISCOMP_ROOT/share/gds/web/ folder to a temporary location prior to the installation of the new GDS package and later on to restore your $SEISCOMP_ROOT/share/gds/web/gds/settings.py and $SEISCOMP_ROOT/share/gds/web/gds/settings-production.py while merging the new configuration options found in the $SEISCOMP_ROOT/share/gds/web/gds/example-settings* files . If you do not want to move your web folder or if you already extracted the GDS package then please remove at least the following files:

rm -rf $SEISCOMP_ROOT/share/gds/web/{django,django_extensions,bitfield}
rm $SEISCOMP_ROOT/share/gds/web/gds/templatetags/breadcrumbs.py

Additional Dependencies

The following new dependencies have been introduced:

  • Python >=3.6

  • Django-3.1

Please refer to Python and Django respectively for installation instructions.

Back-end Configuration

In contrast to previous GDS versions the MIME type of a filter is no longer guessed but must be specified in the GDS back-end configuration, see

Common MIME types are

  • text/gds - GDS bulletin format used by most of the sample filter scripts shipped with GDS

  • text/x-python - Python dictionary displayed as key-value pairs

  • image/[subtype] - Images of any type

  • text/[subtype] - Generic text

  • application/pdf - PDF document

Filter and Spooler Library

The GDS Python library ($SEISCOMP_ROOT/share/gds/tools) uses the seiscomp-python shebang to determine the Python version the SeisComP Python bindings are compiled for. For SeisComP versions <= 2018.327 an alias pointing to either python2 or python3 needs to be created. When the GDS is installed via gsm this will be done automatically. It is recommended to use the same shebang in custom filter and spooler scripts.

Further notable changes are:

  • Renamed $SEISCOMP_ROOT/share/gds/tools/lib/xml.py to $SEISCOMP_ROOT/share/gds/tools/lib/scml.py to avoid naming conflicts

  • Logging facility

    • New messageLimit configuration parameter

    • Try to read log directory from SeisComP Environment, use $HOME/.seiscomp/log as fallback

Migration to 2017.269 (Django 1.11.15)

With version 2017.269 the GDS front-end switched from Django 1.4 to 1.11.15 special attention must be giving to this update.

Additional Dependencies

The following new dependencies have been introduced:

  • Python >=2.7

  • pytz

  • django-1.11.15

  • django-extensions (optional)

  • django-bitfield

Refer to Python and Django for installation instructions.

Database

If you are updating from a GDS version older than 2017.079 make sure to run all appropriate database update scripts to be found in migrations/pre-2017.269

cd $SEISCOMP_ROOT/share/gds/web/gds/migrations/pre-2017.269/mysql
mysql -u sysop -p gds < 2016....
mysql -u sysop -p gds < 2017.079

Although the GDS datamodel was not changed between 2017.079 and 2017.269 you still need to be migrate your database to update the Django system tables. Since your system tables already have been initialized the initial migration step needs to be skipped:

cd $SEISCOMP_ROOT/share/gds/web
python3 manage.py migrate --fake-initial

New Directory Structure

  • All GDS related files are located under the gds subfolder

  • Templates moved from contrib to templates folder

  • CSS and java script files moved from media to static folder, no need to serve media directory with NGINX anymore

Database Backup and Restore

It is a good idea create database backups on regular basis. Especially before upgrading to a newer GDS version a backup is recommended. The specific procedure depends on the database backend chosen:

You may also consider the Django Fixtures as a database independent backup and restore mechanism providing a way to switch database backends and to keep different GDS instances in sync.

MariaDB

Backup

Create a backup of the entire GDS database:

mysqldump -u sysop -p gds > /tmp/gds.sql

Create a backup of the GDS database but omit the rather large GDS log tables:

mysqldump -u sysop -p --no-data gds > /tmp/gds.sql
mysqldump -u sysop -p --no-create-info gds --ignore-table=gds.gds_log_event --ignore-table=gds.gds_log_service --ignore-table=gds.gds_log_recv >> /tmp/gds.sql

Restore

mysql -u sysop -p gds < /tmp/gds.sql

PostgreSQL

Backup

Create a backup of the entire GDS database:

pg_dump -U sysop -W -F c gds > /tmp/gds.sql

Create a backup of the GDS database but omit the rather large GDS log tables:

pg_dump -U sysop -W -F c --exclude-table-data=gds_log_* gds > /tmp/gds.sql

Restore

pg_restore -U sysop -W -c --if-exists -d gds /tmp/gds.sql

Django Fixtures

Django Fixtures are created with the manage.py command dumpdata and restored with the command loaddata.

The GDS datamodel implements support for Natural Keys allowing to dump data with string identifiers instead of numeric database ids. This is in particular useful if you want to keep 2 different database in sync. To take advantage of this feature you have to use the dumpdata options --natural-primary and --natural-foreign.