Rethinking Autoscaling for Apache Solr using Kubernetes
summary
Shows how Kubernetes HPA can drive SolrCloud scaling by extending the Solr Operator to call Solr APIs: it uses the replace node API to safely drain pods on scale‑down, a new balancedReplica API (merged for Solr 9.3) to redistribute shards on scale‑up, and cluster operation locks to avoid concurrent ops. Demo autoscaled from 3→7 nodes for a 12‑replica collection with an HPA target 20% CPU, redistributing replicas successfully and then scaling down one node at a time without data loss.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:06What is Auto scaling
- 5:44Using HPA for Solr
- 10:23Making Solr stateless
- 14:15Ensuring data integrity
- 16:22Live demo setup
- 24:27How replica migration works
- 32:10Balancing replicas logic
- 33:56Project status and thanks
- 34:17Q&A session
related talks
29mScaling Search Clusters with Apache Solr and Kubernetes
40mRunning Solr within Kubernetes at Scale
51mHorizontal Autoscaling with Kubernetes
14mA Tale of Autoscale: Increasing Resiliency while Reducing On-Call Toil
35mSIG Autoscaling Intro
37mBreak Through Cluster Boundaries to Autoscale Workloads Across Them on...
38m50 Shades of K8s Autoscaling - Mario Leander Reimer
35mAutoscaling Kubernetes Deployments: A (Mostly) Practical Guide