Zero downtime maintenance

Hi Dremians,
can you clarify, what operations can be managed with no downtime?
I am interested in Kubernetes deployment.
Is it possible to redeploy K8S worker node containing Dremio PODs and always face to zero downtime?
Is it possible to upgrade version of Dremio without any downtime (rolling upgrade)?
If short downtime is required, do you provide any recommendations how to make resilient an application running on top of Dremio? (retry mechanisms, …)




  • When you redeploy worker nodes, active running queries will get cancelled, increasing worker nodes should not cause any running queries to fail. Decreasing workers may cause some queries to get cancelled
  • Currently, Dremio does not offer rolling upgrade
  • Reflections can have a retry limit set, not user queries, so the application would have to retry once Dremio is up again