treeherder/docs/pulseload.md

Loading Pulse data

For ingestion from Pulse exchanges, on your local machine, you can choose to ingest from any exchange you like. Some exchanges will be registered in settings.py for use by the Treeherder servers. You can use those to get the same data as Treeherder. Or you can specify your own and experiment with posting your own data.

The Simple Case

If you just want to get the same data that Treeherder gets, then you have 3 steps:

1. Create a user on Pulse Guardian if you don't already have one
2. Create your PULSE_URL string
3. Open a Vagrant terminal to read Pushes
4. Open a Vagrant terminal to read Jobs
5. Open a Vagrant terminal to run Celery

1. Pulse Guardian

Visit Pulse Guardian, sign in, and create a Pulse User. It will ask you to set a username and password. Remember these as you'll use them in the next step. Unfortunately, Pulse doesn't support creating queues with a guest account, so this step is necessary.

2. Environment Variable

If your Pulse User was username: foo and password: bar, your config string would be:

PULSE_URL="amqp://foo:bar@pulse.mozilla.org:5671/?ssl=1"

3. Read Pushes

.. note:: Be sure your Vagrant environment is up-to-date.  Reload it and run ``vagrant provision`` if you're not sure.

ssh into Vagrant, then set your config environment variable:

export PULSE_URL="amqp://foo:bar@pulse.mozilla.org:5671/?ssl=1"

Next, run the Treeherder management command to read Pushes from the default Pulse exchange:

./manage.py read_pulse_pushes

You will see a list of the exchanges it has mounted to and a message for each push as it is read. This process does not ingest the push into Treeherder. It adds that Push message to a local Celery queue for ingestion. They will be ingested in step 5.

4. Read Jobs

As in step 3, open a Vagrant terminal and export your PULSE_URL variable. Then run the following management command:

./manage.py read_pulse_jobs

You will again see the list of exchanges that your queue is now mounted to and a message for each Job as it is read into your local Celery queue.

5. Celery

Open your next Vagrant terminal. You don't need to set your environment variable in this one. Just run Celery:

celery -A treeherder worker -B --concurrency 5

That's it! With those processes running, you will begin ingesting Treeherder data. To see the data, you will need to run the Treeherder UI and API. See Running the unminified UI with Vagrant for more info.

Advanced Configuration

Changing which Data to Ingest

treeherder.services.pulse.sources provides default sources for both Jobs and Pushes.

Pushes

push_sources defines a list of exchanges with routing keys. It's rare you'll need to change this so it's not configurable via the environment. However if you wanted to, say, only get pushes to GitHub you would edit the list to look like this:

push_sources = [
    "exchange/taskcluster-github/v1/push.#",
]

Jobs

Job Exchanges and Projects are defined in job_sources, however can also be configured in the environment like so:

PULSE_JOB_SOURCES defines a list of exchanges with projects.

export PULSE_JOB_SOURCES="exchange/taskcluster-treeherder/v1/jobs.mozilla-central:mozilla-inbound,exchange/fxtesteng/jobs.#",

In this example we've defined two exchanges:

• exchange/taskcluster-treeherder/v1/jobs
• exchange/fxtesteng/jobs

The taskcluster-treeherder exchange defines two projects:

• mozilla-central
• mozilla-inbound

The fxtesteng exchange defines a wildcard (#) for its project.

When Jobs are read from Pulse and added to Treeherder's celery queue we generate a routing key by prepending #. to each project key.

Advanced Celery options

If you only want to ingest the Pushes and Jobs, but don't care about log parsing and all the other processing Treeherder does, then you can minimize the Celery task. You will need:

celery -A treeherder worker -B -Q pushlog,store_pulse_jobs,store_pulse_resultsets --concurrency 5

• The pushlog queue loads up to the last 10 Mercurial pushes that exist.
• The store_pulse_resultsets queue will ingest all the pushes from the exchanges specified in push_sources. This can be Mercurial and Github
• The store_pulse_jobs queue will ingest all the jobs from the exchanges specified in job_sources (or PULSE_JOB_SOURCES).

.. note:: Any job that comes from **Pulse** that does not have an associated push will be skipped.
.. note:: It is slightly confusing to see ``store_pulse_resultsets`` there.  It is there for legacy reasons and will change to ``store_pulse_pushes`` at some point.

Posting Data

To post data to your own Pulse exchange, you can use the publish_to_pulse management command. This command takes the routing_key, connection_url and payload_file. The payload file must be a JSON representation of a job as specified in the YML Schema.

Here is a set of example parameters that could be used to run it:

./manage.py publish_to_pulse mozilla-inbound.staging amqp://treeherder-test:mypassword@pulse.mozilla.org:5672/ ./scratch/test_job.json

You can use the handy Pulse Inspector to view messages in your exchange to test that they are arriving at Pulse the way you expect.