ckanext-archiver

Overview

The CKAN Archiver Extension will download all of a CKAN's resources, for three purposes:

1. offer the user it as a 'cached' copy, in case the link becomes broken
2. tell the user (and publishers) if the link is broken, on both the dataset/resource and in a 'Broken Links' report
3. the downloaded file can be analysed by other extensions, such as ckanext-qa or ckanext-pacakgezip.

Demo:

Compatibility: Requires CKAN version 2.1 or later

TODO:

• Show brokenness on the package page (not just the resources)
• Prettify the html bits
• Add brokenness to search facets using IFacet

Operation

When a resource is archived, the information about the archival - if it failed, the filename on disk, file size etc - is stored in the Archival table. (In ckanext-archiver v0.1 it was stored in TaskStatus and on the Resource itself.) This is added to dataset during the package_show call (using a schema key), so the information is also available over the API.

Other extensions can subscribe to the archiver's IPipe interface to hear about datasets being archived. e.g. ckanext-qa will detect its file type and give it an openness score, or ckanext-packagezip will create a zip of the files in a dataset.

Archiver works on Celery queues, so when Archiver is notified of a dataset/resource being created or updated, it puts an 'update request' on a queue. Celery calls the Archiver 'update task' to do each archival. You can start Celery with multiple processes, to archive in parallel.

You can also trigger an archival using paster on the command-line.

By default, two queues are used:

1. 'bulk' for a regular archival of all the resources
2. 'priority' for when a user edits one-off resource

This means that the 'bulk' queue can happily run slowly, archiving large quantities slowly, such as re-archiving every single resource once a week. And meanwhile, if a new resource is put into CKAN then it can be downloaded straight away via the 'priority' queue.

Installation

To install ckanext-archiver:

1. Activate your CKAN virtual environment, for example:

   . /usr/lib/ckan/default/bin/activate
   

2. Install the ckanext-archiver and ckanext-report Python packages into your virtual environment:

   pip install -e git+http://github.com/datagovuk/ckanext-report.git#egg=ckanext-report
   pip install -e git+http://github.com/ckan/ckanext-archiver.git#egg=ckanext-archiver
   

3. Install the archiver dependencies:

   pip install -r ckanext-archiver/requirements.txt
   

4. Now create the database tables:

   paster --plugin=ckanext-archiver archiver init --config=production.ini
   paster --plugin=ckanext-report report initdb --config=production.ini
   

4. Add archiver report to the ckan.plugins setting in your CKAN config file (by default the config file is located at /etc/ckan/default/production.ini).

5. Install a Celery queue backend - see later section.

6. Restart CKAN. For example if you've deployed CKAN with Apache on Ubuntu:

   sudo service apache2 reload
   

Upgrade from version 0.1 to 2.x

NB If upgrading ckanext-archiver and use ckanext-qa too, then you will need to upgrade ckanext-qa to version 2.x at the same time.

NB Previously you needed both ckanext-archiver and ckanext-qa to see the broken link report. This functionality has now moved to ckanext-archiver. So now you only need ckanext-qa if you want the 5 stars of openness functionality.

1. Activate your CKAN virtual environment, for example:

   . /usr/lib/ckan/default/bin/activate
   

2. Install ckanext-report (if not already installed)

   pip install -e git+http://github.com/datagovuk/ckanext-report.git#egg=ckanext-report

3. Add report to the ckan.plugins setting in your CKAN config file (it should already have archiver) (by default the config file is located at /etc/ckan/default/production.ini).

4. Also in your CKAN config file, rename old config option keys if you have them:

   • ckan.cache_url_root to ckanext-archiver.cache_url_root
   • ckanext.archiver.user_agent_string to ckanext-archiver.user_agent_string

5. Upgrade the ckanext-archiver Python package:

   cd ckanext-archiver
   git pull
   python setup.py develop
   

6. Create the new database tables:

   paster --plugin=ckanext-archiver archiver init --config=production.ini
   

7. Ensure the archiver dependencies are installed:

   pip install -r requirements.txt
   

8. Install the developer dependencies, needed for the migration:

   pip install -r dev-requirements.txt
   

9. Migrate your database to the new Archiver tables:

   python ckanext/archiver/bin/migrate_task_status.py --write production.ini
   

Migrations post 2.0

Over time it is possible that the database structure will change. In these cases you can use the migrate command to update the database schema.

::

paster --plugin=ckanext-archiver archiver migrate -c <path to CKAN ini file>

This is only necessary if you update ckanext-archiver and already have the database tables in place.

Installing a Celery queue backend

Archiver uses Celery to manage its 'queues'. You need to install a queue back-end, such as Redis or RabbitMQ.

Redis backend

Redis can be installed like this:

sudo apt-get install redis-server

Install the python library into your python environment:

/usr/lib/ckan/default/bin/activate/pip install redis==2.10.1

It must then be configured in your CKAN config (e.g. production.ini) by inserting a new section, e.g. before [app:main]:

[app:celery]
BROKER_BACKEND = redis
BROKER_HOST = redis://localhost/1
CELERY_RESULT_BACKEND = redis
REDIS_HOST = 127.0.0.1
REDIS_PORT = 6379
REDIS_DB = 0
REDIS_CONNECT_RETRY = True

Number of items in the queue 'bulk':

redis-cli -n 1 LLEN bulk

See item 0 in the queue (which is the last to go on the queue & last to be processed):

redis-cli -n 1 LINDEX bulk 0

To delete all the items on the queue:

redis-cli -n 1 DEL bulk

Installing SNI support

When archiving resources on servers which use HTTPS, you might encounter this error:

requests.exceptions.SSLError: [Errno 1] _ssl.c:504: error:14077410:SSL routines:SSL23_GET_SERVER_HELLO:sslv3 alert handshake failure

Whilst this could possibly be a problem with the server, it is most likely due to you needing to install SNI support on the machine that ckanext-archiver runs. Server Name Indication (SNI) is for when a server has multiple SSL certificates, which is a relatively new feature in HTTPS. This requires installing a recent version of OpenSSL plus the python libraries to make use of this feature.

If you have SNI support installed then this should command run without the above error:

python -c 'import requests; requests.get("http://files.datapress.com")'

On Ubuntu 12.04 you can install SNI support by doing this:

sudo apt-get install libffi-dev
. /usr/lib/ckan/default/bin/activate
pip install 'cryptography==0.9.3' pyOpenSSL ndg-httpsclient pyasn1

You should also check your OpenSSL version is greater than 1.0.0:

python -c "import ssl; print ssl.OPENSSL_VERSION"

Apparently SNI was added into OpenSSL version 0.9.8j but apparently there are reported problems with 0.9.8y, 0.9.8zc & 0.9.8zg so 1.0.0+ is recommended.

For more about enabling SNI in python requests see:

• https://stackoverflow.com/questions/18578439/using-requests-with-tls-doesnt-give-sni-support/18579484#18579484
• https://github.com/kennethreitz/requests/issues/2022

Config settings

1. Enabling Archiver to listen to resource changes

   If you want the archiver to run automatically when a new CKAN resource is added, or the url of a resource is changed, then edit your CKAN config file (eg: development.ini) to enable the extension:

   ckan.plugins = archiver
   

   If there are other plugins activated, add this to the list (each plugin should be separated with a space).

   Note: You can still run the archiver manually (from the command line) on specific resources or on all resources in a CKAN instance without enabling the plugin. See section 'Using Archiver' for details.

2. Other CKAN config options

   The following config variable should also be set in your CKAN config:

   • ckan.site_url = URL to your CKAN instance

   This is the URL that the archive process (in Celery) will use to access the CKAN API to update it about the cached URLs. If your internal network names your CKAN server differently, then specify this internal name in config option: ckan.site_url_internally

3. Additional Archiver settings

   Add the settings to the CKAN config file:

   • ckanext-archiver.archive_dir = path to the directory that archived files will be saved to (e.g. /www/resource_cache)
   • ckanext-archiver.cache_url_root = URL where you will be publicly serving the cached files stored locally at ckanext-archiver.archive_dir.
   • ckanext-archiver.max_content_length = the maximum size (in bytes) of files to archive (default 50000000 =50MB)
   • ckanext-archiver.user_agent_string = identifies the archiver to servers it archives from
   • ckanext-archiver.verify_https = true/false whether you want to verify https connections and therefore fail if it is specified in the URL but does not verify.

4. Nightly report generation

   Configure the reports to be generated each night using cron. e.g.:

   0 6  * * *  www-data  /usr/lib/ckan/default/bin/paster --plugin=ckanext-report report generate --config=/etc/ckan/default/production.ini
   

5. Your web server should serve the files from the archive_dir.

   With nginx you insert a new location after the ckan one. e.g. here we have configured ckanext-archiver.archive_dir to /www/resource_cache and serve these files at location /resource_cache (i.e. http://mysite.com/resource_cache ):

   server {
       # ckan
       location / {
           proxy_pass http://127.0.0.1:8080/;
           ...
       }
       # archived files
       location /resource_cache {
           root /www/resource_cache;
       }
   

Legacy settings

Older versions of ckanext-archiver put these settings in ckanext/archiver/settings.py as variables ARCHIVE_DIR and MAX_CONTENT_LENGTH but this is no longer available.

There used to be an option DATA_FORMATS for filtering the resources archived, but that has now been removed in ckanext-archiver v2.0, since it is now not only caching files, but is seen as a broken link checker, which applies whatever the format.

Using Archiver

First, make sure that Celery is running for each queue. For test/local use, you can run:

paster --plugin=ckanext-archiver celeryd2 run all -c development.ini

However in production you'd run the priority and bulk queues separately, or else the priority queue will not have any priority over the bulk queue. This can be done by running these two commands in separate terminals:

paster --plugin=ckanext-archiver celeryd2 run priority -c production.ini
paster --plugin=ckanext-archiver celeryd2 run bulk -c production.ini

For production use, we recommend setting up Celery to run with supervisord. apt-get install supervisor and use bin/celery-supervisor.conf as a configuration template.

If you are running CKAN 2.7 or higher, configure job workers instead http://docs.ckan.org/en/2.8/maintaining/background-tasks.html#using-supervisor

An archival can be triggered by adding a dataset with a resource or updating a resource URL. Alternatively you can run:

paster --plugin=ckanext-archiver archiver update [dataset] --queue=priority -c <path to CKAN config>

Here dataset is a CKAN dataset name or ID, or you can omit it to archive all datasets.

For a full list of manual commands run:

paster --plugin=ckanext-archiver archiver --help

Once you've done some archiving you can generate a Broken Links report:

paster --plugin=ckanext-report report generate broken-links --config=production.ini

And view it on your CKAN site at /report/broken-links.

Testing

To run the tests:

1. Activate your CKAN virtual environment, for example:

   . /usr/lib/ckan/default/bin/activate
   

2. If not done already, install the dev requirements:

   (pyenv)~/pyenv/src/ckan$ pip install ../ckanext-archiver/dev-requirements.txt
   

3. From the CKAN root directory (not the extension root) do:

   (pyenv)~/pyenv/src/ckan$ nosetests --ckan ../ckanext-archiver/tests/ --with-pylons=../ckanext-archiver/test-core.ini
   

Translations

To translate plugin to a new language (ie. "pl") run python setup.py init_catalog -l pl.

To update template file with new translation added in the code or templates run python setup.py extract_messages in the root plugin directory. Then run ./ckanext/archiver/i18n/unique_pot.sh -v to strip other plugin's translations.

To update translation files for locale "pl" with new template run python setup.py update_catalog -l pl.

Building Debian package

NB this attempt at creating a Debian package is experimental. Important package dependencies have yet to specified. The outstanding issue is that some dependencies do not exist as debian packages (eg: messytables).

To build the debian package:

cd ckanext-archiver; dpkg-buildpackage -us -uc -i -I -rfakeroot

To list the package contents:

dpkg --contents ../python-ckanext-archiver_0.1-1_all.deb

Questions

The archiver information is not appearing on the resource page

Check that it is appearing in the API for the dataset - see question below.

The archiver information is not appearing in the API (package_show)

i.e. if you browse this path on your website: /api/action/package_show?id=<package_name> then you don't see the archiver key at the dataset level or resource level.

Check the paster archiver update command completed ok. Check that the paster celeryd2 run has done the archiving ok. Check the dataset has at least one resource. Check that you have archiver in your ckan.plugins and have restarted CKAN.

'SSL handshake' error

When archiving resources on servers which use HTTPS, you might encounter this error:

requests.exceptions.SSLError: [Errno 1] _ssl.c:504: error:14077410:SSL routines:SSL23_GET_SERVER_HELLO:sslv3 alert handshake failure

This is probably because you don't have SNI support and requires installing OpenSSL - see section "Installing SNI support".