Operating Cassandra in Kubernetes with Temporal – Christian Mauduit & Steven Seidman, Datadog
Operating Cassandra in Kubernetes with Temporal – Christian Mauduit & Steven Seidman, Datadog
Kubernetes support for persistent volumes and statefulsets has made it viable to run stateful applications like Cassandra on Kubernetes. However, it can be challenging to fit every type of operational automation into the narrow customization interfaces that Kubernetes natively offers. At Datadog, we’ve long been running Cassandra on Kubernetes using a combination of init-containers, entrypoint scripts, and pod-lifecycle hooks to automate node addition and replacement. More complex operations continued to require extensive human orchestration, though. Datadog now employs Temporal Workflows to close this gap and automate many ad-hoc operations including infrastructure migrations, progressive rollouts, deployments with extra health-checks and guardrails, and the orchestration of slow processes involving streaming or compaction. We’ll discuss the features our existing tools provide natively and the gaps that they left unaddressed. We’ll talk through the properties of temporal that made it an attractive option to address those gaps, and some of the specific use-cases we’ve implemented as Temporal Workflows.
by The Linux Foundation
linux foundation
