Containing an Elephant: How we moved Hadoop/HBase into Kubernetes and Public Cloud
summary
Salesforce migrates multi‑petabyte HBase/HDFS clusters to Kubernetes on public cloud, keeping performance with EBS and spreading data across three AZs for replica placement. Uses StatefulSets for stable hostnames, patches Hadoop for DNS churn, and extends ExternalDNS for reverse PTRs. Builds a custom controller to batch stateful upgrades and avoid leader churn, implements a keymaker/keytab agent for Kerberos but moves many clusters to Envoy+Istio mTLS for simpler rotation and fewer DNS/Perf failures, prototype in 3–4 month prototype, production readiness in ~8 months.
AI-written summary · about
chapters (17)
- 0:00Introduction to talk
- 1:12Why Kubernetes for Hadoop/HBase
- 3:19Immutable deployments in Kubernetes
- 5:43Scalability and high availability
- 8:13Storage in Kubernetes
- 10:21Consistency across clouds
- 11:47Challenges with stateful apps
- 13:52StatefulSets for Hadoop
- 16:00Custom controllers for upgrades
- 18:08Pod disruption budgets
- 19:34Leader election issues
- 20:57DNS challenges in public cloud
- 24:12Global DNS replication
- 26:57Availability zones in cloud
- 29:47Authentication and encryption
- 35:04Transition to TLS and Envoy
- 38:13Summary and hiring
related talks
24mDynamic Big Data in the Cloud with Kubernetes
38mChallenges of and Solutions for Migrating Spark From Legacy Hadoop Clusters To Kubernetes
31mA Journey of Migrating Data Platform from Data Center to Cloud
26mRe-architecting the Data Platform with Spark on Kubernetes
45mHadoop in the Cloud: The Story of Migrating Hundreds of Petabytes
53mBig Data: from mammoth to elephant
33mContainerizing & Orchestrating Hadoop Infrastructure at Scale
14mKeynote: The Big Leap: Conquering the Legacy System and Building a Scalable Cloud Infra...