Hi everyone 👋
I'm looking for guidance on deploying Cassandra and OpenSearch in a Kubernetes environment connected to a DMA cluster.
Specifically, I'd appreciate help with:
- Any documentation, best practices, or reference architectures for this setup.
- Cost considerations across different deployment scenarios (a customer needs this to evaluate options).
- Recommendations or warnings against pursuing the Kubernetes approach, if there are known pitfalls or reasons to avoid it, that feedback would be extremely valuable too.
Thanks so much in advance for your support! 🙏
Best regards
Hi Jardel,
DataMiner itself is agnostic as to the deployment method of the storage. So from a DataMiner configuration point of view, all is as you're currently used to when it comes to connecting the DMS to the storage. As long as the DMS can reach the nodes of the Cassandra & OpenSearch clusters then it's good to go.
So that brings us to the Kubernetes. We do not currently provide a Kubernetes-specific reference architecture and have limited experience deploying these databases on Kubernetes ourselves. For Cassandra, K8ssandra is a well-known community project providing Kubernetes operators and supporting components for deploying and operating Cassandra clusters. Similarly, OpenSearch provides their own Kubernetes Operator.
Pitfalls wise, the main things to keep in mind is to ensure that the deployed containerized Cassandra & OpenSearch clusters follow the recommendations from out bare-metal install guide as far as the cluster settings go. This mainly pertains to settings such as tombstone warning thresholds, query clause counts, timeout settings, etc... As these are more closely aligned with the behaviour of the DataMiner software rather than the infrastructural side of the storage deployment.
Relevant references:
In short, Kubernetes is technically possible from DataMiner’s perspective, but like any on-prem infrastructure it is an additional responsibility to operate. Our own tooling and experience are primarily focused on traditional deployments, but for the best experience we recommend STaaS to shed yourself of the worries of storage management.