Job Metadata & Run Tags#

Metadata and tags provide two different ways for you to attach information to your jobs and the runs launched from those jobs. The main difference between metadata and tags is what "object" the information is attached to. Metadata is attached to JobDefinitions (specified using the @job decorator), and tags are attached to runs that are created by executing a job.

Metadata#

Metadata allows you to attach information to a job. This information can be whatever you want, but possible cases include:

keeping track of the team responsible for maintaining a job
linking to documentation or other resources
displaying the git hash corresponding to the current job definition

Note: If you are running Dagster using separate webserver and user code installations (more info here), then your dagster-webserver installation must be >=0.14.18 to use metadata on jobs.

Specifying metadata#

When you attach metadata to a job, you do it as a dictionary of key value pairs. The keys must be a string, but the values can be any one of the MetadataValue classes we provide. You can also use primitive python types as values, and dagster will convert them to the appropriate MetadataValue.


@op
def my_op():
    return "Hello World!"


@job(
    metadata={
        "owner": "data team",  # will be converted to MetadataValue.text
        "docs": MetadataValue.url("https://docs.dagster.io"),
    }
)
def my_job_with_metadata():
    my_op()

In addition to adding metadata on the @job decorator, you can also add metadata using the GraphDefinition.to_job method.


@graph
def my_graph():
    my_op()


my_second_job_with_metadata = my_graph.to_job(
    metadata={"owner": "api team", "docs": MetadataValue.url("https://docs.dagster.io")}
)

Viewing Metadata#

After attaching metadata to a job, you can view it in the Dagster UI by navigating to the job overview page. Metadata will be displayed in the right pane.

Tags#

Tags allow you to attach information to the run created when you execute a job. Tags can contain any information you want, and dagster will also attach some tags to your runs (we'll cover these later).

Specifying tags#

You can specify tags you want attached to every run by adding them to a job. Tags are specified as a dictionary of key value pairs where the key must be a string and the value must be a string or json that is serializable to a string.


@job(tags={"foo": "bar"})
def my_job_with_tags():
    my_op()

In addition to adding tags on the @job decorator, you can also add metadata using the GraphDefinition.to_job method.


my_second_job_with_tags = my_graph.to_job(tags={"foo": "bar"})

When executing a job, you can add tags to the run using the Launchpad in the UI

Viewing Tags#

You can view the tags that have been attached to runs by going to the Runs page in the UI

Dagster-provided tags#

Dagster will automatically add tags to your runs in some cases including:

dagster/op_selection - The op selection for the run, if applicable
dagster/partition - The partition of the run, if applicable
dagster/schedule_name - The schedule that triggered the run, if applicable
dagster/sensor_name - The sensor that triggered the run, if applicable
dagster/backfill - The backfill ID, if applicable
dagster/parent_run_id - The parent run of a re-executed run
dagster/image - The Docker image tag

Using tags to affect run execution#

Some features of Dagster are controlled using the tags attached to a run. Some examples include:

You are viewing an outdated version of the documentation.