Batch job and queues can be done in k8s nowadays See also: Vulcano, Armada.
Running Apache Kafka at scale on Kubernetes with Strimzi
Apache Kafka on K8s - https://strimzi.io/ OpenAPI schema for ordering Kafka topic, a db, etc. DevEx -> Orchestration -> Shared services -> Resources
Platform Real State: Abstract Your Organization's Tenancy Model Away With Crossplane
Saving costs, sharing infra, reducing noisy neighbour. Check:
- What infra and app services tenants need
- Inventorizing - find out what is required
- App devs need
- GitOps, auth, secrets, DNS, expose endpoints
- Keycloak, external-dns, ArgoCD, ingress-nginx
- Platform team needs
- Create stuff outside of k8s, DNS, isolation, RBAC, GitOps
- Kyverno, Crossplane
- What scale
- Difference between small and large tenants
- Limitations due to resources you use (LBs, number of pods)
- Dedicated/shared
- Isolation
- Compute, network, control panel, middleware (tools, app services)
- How much k8s is visible
- Simplifies GitOps on non-k8s resources
- Custom CRD that masquerades as a Helm chart
- Efficiency *
Saxo
Create blueprints for services Reduce number of approvals you need Developers like GitOps - restore entire cluster from git Blueprint components:
- Everything is in YAML
- CRDs and operators Backstage format for blueprints
Operator is in .NET https://github.com/buehler/dotnet-operator-sdk?
Meta: what is the point of reporting "this piece is missing" when using declarative infra? Shouldn't everything be reconciled anyway and minimize the number of such errors?