🔖 Feature description
Add three enhancements to the Devtron reference Helm chart (scripts/devtron-reference-helm-charts), shipped as a new chart version.
priorityClassName support on the Rollout's pod spec, so an existing Kubernetes PriorityClass can be assigned to application pods for scheduling priority and preemption.
service.trafficDistribution support on the generated Service, defaulting to nil (unset, standard routing), with allowed values PreferSameZone, PreferSameNode and PreferClose (Kubernetes' topology-aware traffic routing field).
A default resources.requests.ephemeral-storage of 1Gi and resources.limits.ephemeral-storage of 5Gi, so workloads get a predictable local-storage footprint out of the box. cpu and memory remain unset, unchanged.
🎤 Pitch / Usecases
These are common production-hardening knobs that today require hand-editing the generated manifests or a raw YAML patch.
priorityClassName lets platform teams ensure critical workloads preempt lower-priority pods under node pressure.
service.trafficDistribution lets teams reduce cross-zone traffic and latency by preferring same-zone or same-node routing on clusters that support it.
A sane ephemeral-storage default protects nodes from disk-pressure evictions caused by workloads with no storage ceiling, without imposing any default cpu/memory limits.
🔄️ Alternative
No response
👀 Have you spent some time to check if this issue has been raised before?
🏢 Have you read the Code of Conduct?
🔖 Feature description
Add three enhancements to the Devtron reference Helm chart (scripts/devtron-reference-helm-charts), shipped as a new chart version.
priorityClassName support on the Rollout's pod spec, so an existing Kubernetes PriorityClass can be assigned to application pods for scheduling priority and preemption.
service.trafficDistribution support on the generated Service, defaulting to nil (unset, standard routing), with allowed values PreferSameZone, PreferSameNode and PreferClose (Kubernetes' topology-aware traffic routing field).
A default resources.requests.ephemeral-storage of 1Gi and resources.limits.ephemeral-storage of 5Gi, so workloads get a predictable local-storage footprint out of the box. cpu and memory remain unset, unchanged.
🎤 Pitch / Usecases
These are common production-hardening knobs that today require hand-editing the generated manifests or a raw YAML patch.
priorityClassName lets platform teams ensure critical workloads preempt lower-priority pods under node pressure.
service.trafficDistribution lets teams reduce cross-zone traffic and latency by preferring same-zone or same-node routing on clusters that support it.
A sane ephemeral-storage default protects nodes from disk-pressure evictions caused by workloads with no storage ceiling, without imposing any default cpu/memory limits.
🔄️ Alternative
No response
👀 Have you spent some time to check if this issue has been raised before?
🏢 Have you read the Code of Conduct?