The 'jobs' view

The private tables that jobs are actually stored into are unstable: we may change them in a patch release, so you should not use them. As of Graphile Worker 0.16, the graphile_worker.jobs view exists as a stable interface to let you see details of enqueued jobs. The jobs view may gain additional columns over time, but any column deletions or type changes will require a semver major release of Graphile Worker.

Performance considerations

danger

You should not read from the jobs view frequently; any reading from Graphile Worker's tables can cause a performance impact on the running workers, and doing this too often could cause major performance degradation for you - especially if you access many rows, or your read does not use an index.

When reading from the jobs view, it's recommended that you only select the columns you truly need, and that you apply efficient filters to ensure that Postgres looks at the fewest number of jobs possible.

danger

Do not read from the jobs view from within a transaction; this could cause performance issues!

Columns

• id - the primary key of the job
• queue_name - the name of the queue (if any) this job was added to
• task_identifier - the identifier of the task this job wants to execute
• priority - the "priority" (really the "nice") of the job; a numerically lower (including negative) value indicates the job should execute before tasks with a numerically higher value
• run_at - when the job is scheduled to run
• attempts - how many times we've attempted to execute this job
• max_attempts - the maximum number of times we'll attempt this job
• last_error - if an error occurred the last time this job was executed, what the error was
• created_at - when the job was inserted into the database
• updated_at - when the job was last updated
• key - the job_key of the job, if any
• locked_at - when the job was locked, if locked
• locked_by - the worker id that the job was locked by, if locked
• revision - the revision number of the job, bumped each time the record is updated
• flags - the forbidden flags associated with this job

info

The job payload is deliberately not included in the jobs view to avoid people from performing expensive filtering using it. If you need to see the payload of a job, you should use a tracking table instead. If you need it for debugging then you can read it from the private tables, just be careful, and don't write scripts to do it for you since it might change in a patch release.