Support tasks for CKAN
CKAN is the publishing application for data.gov.uk. ckanext-datagovuk is the primary CKAN extension for data.gov.uk.
Environments
There are three environments for CKAN:
You can SSH onto these machines in same way as all other GOV.UK AWS applications. The machine node class is ckan.
Managing CKAN
First check to see if it is possible to complete the task through the web interface
(credentials are available in the govuk-secrets password store, under datagovuk/ckan).
For commands not available via the user interface you must connect to the server to perform these
tasks. Most of the commands to interact with CKAN use a tool called paster. Many of these
commands take a path to the config file with the -c option, which is located at /var/ckan/ckan.ini
in our deployments.
On GOV.UK servers paster should be run with:
cd /var/apps/ckan
sudo -u deploy govuk_setenv ckan venv/bin/paster [COMMAND] -c /var/ckan/ckan.ini
Initialising the database
There may be times when you need to start with an empty database (e.g. on integration). The following commands will create the relevant schema for core CKAN and the harvesting extension on integration.
. /var/apps/ckan/venv/bin/activate
paster --plugin=ckan db init -c /var/ckan/ckan.ini
paster --plugin=ckanext-harvest harvester initdb -c /var/ckan/ckan.ini
Accessing the database
In order to access the CKAN database to run queries on the db_admin machine:
psql -U ckan -h postgresql-primary -p 5432 ckan_production
The password can be extracted from the configuration file on the ckan machine for the environment you are targeting:
more /var/ckan/ckan.ini | grep sqlalchemy
Accessing the CKAN API
There are times when it can be useful to access the CKAN API when debugging or resolving issues. Note that the responses will be different depending on your access permissions.
Queries can use the package ID or name (the slug) e.g.
https://data.gov.uk/api/3/action/package_search?q=id:93a39f01-7bba-430f-aa35-c30bf2d88b2f
returns the same as
https://data.gov.uk/api/3/action/package_search?q=name:north-lincolnshire-brown-field-register
Here are some more complex examples of using the API, as well as some where the documentation is more sparse:
# Retrieve full details about a package (dataset)
https://data.gov.uk/api/3/action/package_search?q=name:north-lincolnshire-brown-field-register
# Find all packages created during a specific timeframe
https://data.gov.uk/api/3/action/package_search?q=metadata_created:[2017-06-01T00:00:00Z%20TO%202017-06-30T00:00:00Z]
# Find all packages modified during a specific timeframe
https://data.gov.uk/api/3/action/package_search?q=metadata_modified:[2017-06-01T00:00:00Z%20TO%202017-06-30T00:00:00Z]
# Count all packages and list a sample
https://ckan.publishing.service.gov.uk/api/search/dataset?fl=id,dataset_type,name,title,extras_guid,extras_harvest_source_title,extras_harvest_object_id&q=state:active%20and%20dataset_type:dataset&limit=20
# Count harvested packages and list a sample
https://ckan.publishing.service.gov.uk/api/search/dataset?fl=id,state,dataset_type,name,title,extras_guid,extras_harvest_source_title,extras_harvest_object_id&q=state:active%20and%20extras_harvest_object_id:[%22%22%20TO%20*]%20and%20dataset_type:dataset&limit=20
# Count manually published packages and list a sample
https://ckan.publishing.service.gov.uk/api/search/dataset?fl=id,dataset_type,name,title,extras_guid,extras_harvest_source_title,extras_harvest_object_id&q=state:active%20and%20-extras_harvest_object_id:*%20and%20dataset_type:dataset&limit=20
# Count packages harvested in last 7 days and list a sample https://ckan.publishing.service.gov.uk/api/search/dataset?fl=id,state,dataset_type,name,title,extras_guid,extras_harvest_source_title,extras_harvest_object_id&q=state:active%20and%20extras_harvest_object_id:[%22%22%20TO%20*]%20and%20dataset_type:dataset%20and%20metadata_modified:[NOW-7DAY/DAY%20TO%20NOW]&limit=20
# Count packages manually published in last 7 days and list a sample https://ckan.publishing.service.gov.uk/api/search/dataset?fl=id,dataset_type,name,title,extras_guid,extras_harvest_source_title,extras_harvest_object_id&q=state:active%20and%20-extras_harvest_object_id:*%20and%20dataset_type:dataset%20and%20metadata_modified:[NOW-7DAY/DAY%20TO%20NOW]&limit=20
# Count packages from Daily harvest sources and list a sample
https://ckan.publishing.service.gov.uk/api/search/dataset?fl=id,metadata_modified,dataset_type,name,title,extras_guid,extras_harvest_source_title,extras_harvest_object_id,extras_frequency-of-update&q=state:active%20and%20extras_harvest_object_id:%5B%22%22%20TO%20*%5D%20and%20frequency-of-update:daily%20and%20dataset_type:dataset&limit=20
# List all publishers
https://data.gov.uk/api/3/action/organization_list
# View a publisher record
https://data.gov.uk/api/3/action/organization_show?id=government_digital_service
# View a user (e.g. to get CKAN API key for a publishing user)
https://data.gov.uk/api/3/action/user_show?id=user_d484581
Creating a system administrator account
paster --plugin=ckan sysadmin add USERNAME email=EMAIL_ADDRESS -c /var/ckan/ckan.ini
You will be prompted twice for a password.
Removing a system administrator account
paster --plugin=ckan sysadmin remove USERNAME -c /var/ckan/ckan.ini
Managing users
Listing users
paster --plugin=ckan user list -c /var/ckan/ckan.ini
Viewing a user
paster --plugin=ckan user USERNAME -c /var/ckan/ckan.ini
Adding a user
paster --plugin=ckan user add USERNAME email=EMAIL_ADDRESS -c /var/ckan/ckan.ini
Removing a user
paster --plugin=ckan user remove USERNAME -c /var/ckan/ckan.ini
Changing a user’s password
paster --plugin=ckan user setpass USERNAME -c /var/ckan/ckan.ini
Managing publishers
Change a publisher’s name
Change the name in the publisher page then reindex that publisher:
paster --plugin=ckan search-index rebuild-publisher [PUBLISHER] -c /var/ckan/ckan.ini
Managing datasets
Deleting a dataset
CKAN has two types of deletions, the default soft-delete, and a purge. The soft delete gives the option of undeleting a dataset but the purge will remove all trace of it from the system.
Where the following commands mention DATASET_NAME, this should either be the slug for the dataset, or the UUID.
Deleting a dataset:
- Find the dataset in the CKAN UI
- Click on the ‘Manage’ button, then the ‘Delete’ button at the bottom of the page
The ‘Delete’ button is currently not available for draft datasets. In order to soft-delete a draft dataset, follow the above steps, but manually change ‘edit’ to ‘delete’ on the ‘Manage’ page for the dataset.
Purging a dataset:
paster --plugin=ckan dataset purge DATASET_NAME -c /var/ckan/ckan.ini
There may be times when a large number of datasets must be deleted. This can be done remotely from your machine using the CKAN API. Your API key is required, which can be obtained from your user profile on the web interface. Put a list of dataset slugs or GUIDs in a text file, with one dataset per line, then run the following.
while read p; do curl --request POST --data "{\"id\": \"$p\"}" --header "Authorization: <your_api_key>" https://data.gov.uk/api/3/action/package_delete; done < list_of_ids.txt
After deleting or purging a dataset, it will take up to 10 minutes to update on Find, due to the sync process.
Register a brownfield dataset
See the supporting manual.
Rebuilding the search index
CKAN uses Solr for its search index, and occasionally it may be necessary to interact with it to refresh the index, or rebuild it from scratch.
Refresh the entire search index (this adds/removes datasets, but does not clear the index first):
paster --plugin=ckan search-index rebuild -r -c /var/ckan/ckan.ini
Rebuild the entire search index (this deletes the index before re-indexing begins):
paster --plugin=ckan search-index rebuild -c /var/ckan/ckan.ini
Rebuilding the entire search index immediately removes all records from the search before re-indexing begins. No datasets will be served from the
package_searchAPI endpoint until the re-index has completed. This command should therefore only be used as a last resort since it will cause the sync process to assume there is no data for a period of time.
Only reindex those packages that are not currently indexed:
paster --plugin=ckan search-index -o rebuild -c /var/ckan/ckan.ini
csw endpoint
The csw endpoint should be available on https://data.gov.uk/csw which
should redirect to https://ckan.publishing.service.gov.uk/csw.
csw endpoint unavailable
If it is not showing xml with an error Missing keyword: service you can check
that it is running on the ckan machine:
$ sudo service pycsw_web-procfile-worker status
pycsw_web-procfile-worker start/running
$ $ ps aux | grep pycsw
root 29503 0.0 0.0 59652 2040 ? Ss 06:04 0:00 sudo -u deploy -E sh -c PATH=/usr/lib/rbenv/shims:$PATH exec unicornherder --gunicorn-bin ./venv/bin/gunicorn -p /var/run/ckan/pycsw_unicornherder.pid -- ckanext.datagovuk.pycsw_wsgi --bind localhost:${PYCSW_PORT} --timeout ${GUNICORN_TIMEOUT} --workers ${GUNICORN_WORKER_PROCESSES} --log-file /var/log/ckan/pycsw.out.log --error-logfile /var/log/ckan/pycsw.err.log 2>>'/var/log/ckan/procfile_pycsw_web.err.log' 1>>'/var/log/ckan/procfile_pycsw_web.out.log'
...
If no running processes are found then you can restart it using this command:
$ sudo service pycsw_web-procfile_worker restart
It is worth checking the pycsw logs to investigate why it failed:
$ tail -f /var/log/ckan/pycsw.err.log
You can get a summary of csw records available from this url https://ckan.publishing.service.gov.uk/csw?service=CSW&version=2.0.2&request=GetRecords&typenames=csw:Record&elementsetname=brief
Syncing the csw records with ckan datasets
Normally the sync between csw and ckan will start at 6 each day, but in
case it should fail or if the sync needs to happen sooner you can manually
trigger the sync after Solr has been reindexed.
$ paster --plugin=ckanext-spatial ckan-pycsw load -p /var/ckan/pycsw.cfg -u http://localhost:3220
Harvesting
Although harvesters can mostly be managed from the user interface, it is sometimes easier to perform these tasks from the command line.
List current jobs
Returns a list of currently running jobs. This will contain the JOB_ID necessary to cancel jobs.
paster --plugin=ckanext-harvest harvester jobs -c /var/ckan/ckan.ini
It may be faster to run a SQL query to get the ID of a specific harvest job.
You can do this by running the command from Accessing the database
and passing a -c argument:
<psql_ckan_production_command> -c "SELECT id FROM harvest_source WHERE name = '[NAME]'"
Find the harvest source of a dataset
The harvest source of a dataset can be found using the CKAN API. Once you have the slug of the dataset, you can go to the following URL:
https://ckan.publishing.service.gov.uk/api/action/package_show?id=<slug>
In the JSON response there should be a harvest field which will tell you the
ID and name of the harvest source.
Get the status of a harvester
- Login to CKAN as a sysadmin user (credentials are available in the
govuk-secretspassword store, underdatagovuk/ckan). - Navigate to the relevant harvester (use the ‘Harvest’ button in the header).
- You will see a list of the datasets imported by this harvest source.
- Click the ‘Manage’ button to get the status.
- A summary of the current status will be shown. Individual runs (and the error messages logged) can be access from the ‘Jobs’ tab.
Restart a harvest job
- Follow the steps to get the status of a harvester.
- If the harvester is currently running, click the ‘Stop’ button to stop it. Once it has stopped, or if it is not currently running, click the ‘Reharvest’ button. You will know if the harvester is running because the ‘Reharvest’ button will be disabled.
If the harvest job is hanging and the ‘Stop’ button is not responding, you will have to log on to the ckan machine to restart it:
- SSH into the ckan machine with
gds govuk connect -e production ssh ckan - Assume the deploy user -
sudo su deploy - Activate the virtual environment -
. /var/apps/ckan/venv/bin/activate - Run the harvest job manually -
paster --plugin=ckanext-harvest harvester run_test <harvest source> -c /var/ckan/ckan.ini- where
harvest sourceis from the url when visiting the harvest source page, it will be something likecabinet-office
- where
If the job fails to complete the ticket should be updated with comments and prioritised to low for the product owner to review.
Cancel a harvest job
- Follow the steps to get the status of a harvester.
- Click the ‘Stop’ button to stop it.
Sometimes a harvest job can get stuck and not complete, and it’s not possible to cancel it through the UI.
You can get the JOB_ID from the harvest dashboard under “Last Harvest Job” (or from Listing current jobs).
Then cancel the job by running:
paster --plugin=ckanext-harvest harvester job_abort JOB_ID -c /var/ckan/ckan.ini
This can also be done by running SQL:
<psql_ckan_production_command> -c "UPDATE harvest_job SET finished = NOW(), status = 'Finished' WHERE source_id = '[UUID]' AND NOT status = 'Finished';"
Purging all currently queued tasks
It may be necessary, if there is a schedule clash and the system is too busy, to purge the queues used in the various stages of harvesting
WARNING
This command will empty the Redis queues
paster --plugin=ckanext-harvest harvester purge_queues -c /var/ckan/ckan.ini
Restarting the harvest service
The harvesting process runs as a single threaded program, if any harvesting process crashes by raising an exception, it will take out the entire process. We have configured Upstart to restart the process automatically, but if the service keeps crashing, Upstart will decide it’s unhealthy and stop that after a while.
You can check whether the process is still running by checking if entries are
still being written to the log file on the ckan machine:
$ gds govuk connect -e production ssh ckan
$ sudo tail -f /var/log/ckan/procfile_harvester_fetch_consumer.err.log
Or you could check that the services are all showing as started on the ckan
machine:
$ sudo initctl list | grep harvester
If any of the services are not showing as started, you will need to restart
them. There are both ‘gather’ and ‘fetch’ jobs that may need restarting:
‘Gather’ jobs retrieve the identifiers of the updated datasets and create jobs in the fetch queue. To restart, run:
$ sudo initctl restart harvester_gather_consumer-procfile-worker
‘Fetch’ jobs retrieve the datasets from the remote source and perform the relevant updates in CKAN. To restart, run:
$ sudo initctl restart harvester_fetch_consumer-procfile-worker
Alternatively, if the server has stopped, there is a Fabric script that will restart it for you. It will only restart if it detects the harvesting process is no longer running, so is safe to run immediately if you suspect the process has crashed:
$ fab aws_production class:ckan ckan.restart_harvester
CKAN publisher on Staging environment responds with Nginx 504 timeout:
Sometimes CKAN publisher on Staging environment responds with a 504 from Nginx, this is due to it timing out when connecting to the database as there are too many connections, current limit is 1000, though under normal working the number of connections should be under 100 if the system is not under load testing.
SELECT rolname, rolconnlimit FROM pg_roles WHERE rolname='ckan';`
Changing the connection limit
The connection limit can be updated, but should not exceed the hard limit which is around 3000.
ALTER USER ckan WITH CONNECTION LIMIT 1001;
Trying to log on to the database will result in this error:
FATAL: too many connections for role "ckan"
Log in as another user:
sudo psql -U aws_db_admin -h postgresql-primary --no-password -d ckan_production
View the number of connections and types of queries:
SELECT COUNT(*) FROM pg_stat_activity WHERE datname = 'ckan_production';
It has been observed that during an overnight sync one of the queries caused a large number of db connections:
The following query captures part of that query and uses it to target the pid:
SELECT pid FROM pg_stat_activity WHERE datname = 'ckan_production' and query LIKE 'SELECT "user".password AS user_password%';
To cancel these queries
SELECT pg_cancel_backend(pid)
FROM pg_stat_activity
WHERE pid IN
(SELECT pid
FROM pg_stat_activity
WHERE datname = 'ckan_production' and query LIKE 'SELECT "user".password AS user_password%');
To identify long running queries
SELECT
pid,
now() - pg_stat_activity.query_start AS duration,
query,
state
FROM pg_stat_activity
WHERE
(now() - pg_stat_activity.query_start) > interval '5 minutes' AND
datname = 'ckan_production';
Cancel long running queries
This will take a few seconds to be processed
SELECT pg_cancel_backend(pid)
FROM pg_stat_activity
WHERE
(now() - pg_stat_activity.query_start) > interval '5 minutes' AND
state = 'active' AND
datname = 'ckan_production';
If cancelling does not work you can terminate the query
Note - this must be used with caution as may cause the database to restart to recover consistency.
SELECT pg_terminate_backend(pid)
FROM pg_stat_activity
WHERE
(now() - pg_stat_activity.query_start) > interval '5 minutes' AND
state = 'active' AND
datname = 'ckan_production';