Autoscaling FAQ
Review frequently asked questions about Cloudera Data Hub autoscaling.
- A node is not running any tasks, why is it not being scaled down?
- There are multiple possibilities:
- The cluster may already be at min-node-count.
- Check if running on a Cloudera Runtime version prior to 7.2.15, with max-capacity of queues not set to 100%.
- Check user-limit-factor. Is this preventing YARN from executing additional tasks for a user?
- Nodes that are not running containers, but have run containers for a running job, will not be considered for load-based downscaling. This is because the data generated on these nodes could be required while the job is running.
- Will load-based autoscaling remove nodes that have running tasks?
- Not unless the maximum node count limit is reduced. This forces the removal of certain nodes to reach the new limit.
- I've encountered the following error: "Autoscaling Trigger Failed. Autoscaling Collection metrics via user <CRN> failed." How should I proceed?
- For Cloudera accounts with a large number of users, autoscaling might encounter failures when capturing cluster metrics to scale up/down. If you encounter this issue please contact Cloudera Support.