Superset Operator
- Commercial Systems
Channel | Revision | Published | Runs on |
---|---|---|---|
latest/stable | 35 | 02 Oct 2024 | |
latest/beta | 32 | 12 Sep 2024 | |
latest/edge | 35 | 02 Oct 2024 |
juju deploy superset-k8s --channel edge
Deploy Kubernetes operators easily with Juju, the Universal Operator Lifecycle Manager. Need a Kubernetes cluster? Install MicroK8s to create a full CNCF-certified Kubernetes system in under 60 seconds.
Platform:
This guide describes steps you can take to optimise your Superset deployment performance.
Enable asynchronous querying
To enable asynchronous querying, you need to deploy Charmed Superset Workers. These workers handle long-running queries asynchronously, allowing the User Interface (UI) to remain responsive. You can do this as follows:
juju deploy superset-k8s --config charm-function=worker superset-k8s-worker
To be functional, the Charmed Superset Worker requires relations with the same PostgreSQL and Redis charms as the server. These can be added as below:
juju relate superset-k8s-worker postgresql-k8s
juju relate superset-k8s-worker redis-k8s
Using the UI, you can enable Asynchronous Query Execution (AQE) at the database level.
To do so, edit the database. Under Performance
, check the Asynchronous query execution
box.
This is recommended for all production databases to relieve load on the UI.
Enable beat scheduling
Superset’s scheduling system relies on a single instance of the beat scheduler. This scheduler handles periodic jobs like caching or data refreshes. Only one instance should be deployed to avoid conflicting schedules. This can be deployed as follows:
juju deploy superset-k8s --config charm-function=beat superset-k8s-beat
To be functional, the Charmed Superset Beat requires relations with the same PostgreSQL and Redis charms as the server and worker(s). These can be added as below:
juju relate superset-k8s-beat postgresql-k8s
juju relate superset-k8s-beat redis-k8s
Scaling applications
Charmed Superset supports independent scaling of the web server and workers. The web server and workers can be scaled horizontally to handle more load, while the beat scheduler should remain singular.
Use the juju scale-application
command to adjust the number of instances of each service as needed:
juju scale-application superset-k8s -n 3
Three units of server and worker applications for a production deployment are recommended.