Skip to content

Pipeline Scaling Limits in Community Edition

Our free Pachyderm Community Edition contains built-in scaling limitations and parallelism thresholds. If you want to scale beyond these limits, request your FREE 30-Day Pachyderm Enterprise token and enjoy unlimited scaling, JupyterHub integration, and more.

You might qualify for a free Enterprise license.

Pachyderm offers activation keys for proofs-of-concept, startups, academic, nonprofit, or open-source projects. Tell us about your project!

Scaling Limits

Number of concurrent pipelines deployed Number of workers for each pipeline
Community Users can deploy up to 16 pipelines. Community Users can run up to 8 workers in parallel on each pipeline.

What happens when you exceed those limits?

As a general rule, Pachyderm provides an error message in the STDERR whenever a limit is encountered that prevents you from successfully running a command. In that case, the alert message links to a free trial request form.

Limit on the number of pipelines

When exceeding the number of pipelines:

  • pachctl create pipeline fails once the maximum number of pipelines is reached.

  • pachctl update pipeline and pachctl edit pipeline succeed on existing pipelines, fail when attempting to create pipelines beyond the limit.

Note

If update pipeline fails for any other reason, it does not log any message related to pipeline limits.

  • pachctl restore fails when trying to restore with too many pipelines.

All of the commands listed above create a distinct message to STDERR and to the pachd logs. This message includes information such as the limit on the number of pipelines in the Community Edition, the total number of pipelines deployed, and provides a link to request an Enterprise key to lift those limitations.

  • all other extract, list, run, start, stop pipeline commands' behavior remains unchanged.

Limit on the number of workers per pipeline

Pachyderm has two parallelism strategies: constant and coefficient.

When constant parallelism > 8:

  • pachctl create pipeline and pachctl update pipeline fail. A message to STDERR and pachd logs is generated. You will need to update your pipeline specification file accordingly or activate an Enterprise license.

The previous behavior differs for a coefficient parallelism:

  • pachctl create, update, edit pipeline will succeed. However, Pachyderm will scale the number of workers up to the Community Edition limit, and ignore the remaining workers if the value set on coefficient results in a request for more than 8 workers. A message is logged in pachd logs.

What happens when you upgrade to 1.13 and have more than 16 pipelines?

All existing pipelines continue to work after an upgrade to 1.13. pachd logs will mention the successful startup above the "update pipeline message". However, you will not be able to create additional pipelines.

Note

Pipelines automatically generated by the system (for example, build pipelines, cron...) are not considered when assessing the total number of pipelines deployed. The limit applies to user-created pipelines only.


Last update: April 15, 2021
Does this page need fixing? Edit me on GitHub