Kubernetes Resource Management

Recommended Kubernetes resources configuration and management practices for deploying Prisme.ai.

Proper Kubernetes resource configuration ensures optimal performance, reliability, and cost-efficiency for your Prisme.ai deployment. This guide details best practices and recommended configurations.

---

Recommended Resource Configuration

The following guidelines help ensure optimal performance for a standard Prisme.ai deployment:

Component	CPU Requests	CPU Limits	Memory Requests	Memory Limits
API Gateway	500m	1000m	512Mi	1Gi
Runtime	1000m	2000m	1Gi	2Gi
Console (Studio)	250m	500m	256Mi	512Mi
Pages	250m	500m	256Mi	512Mi
Events	500m	1000m	512Mi	1Gi
MongoDB (per node)	1000m	2000m	2Gi	4Gi
Elasticsearch (per node)	1000m	2000m	4Gi	8Gi
Redis (per node)	500m	1000m	1Gi	2Gi

---

Best Practices for Resource Management

Implement these best practices to optimize Kubernetes resource usage:

Resource Requests and Limits

• Set accurate resource requests to inform Kubernetes scheduler for optimal pod placement.
• Configure resource limits to prevent resource starvation.

Autoscaling

• Use Horizontal Pod Autoscaler (HPA) based on CPU/Memory utilization.
• Set minimum and maximum replica counts for critical components like API Gateway and Runtime.

Node Pool Optimization

• Configure separate node pools for CPU-intensive (Runtime) and memory-intensive (Elasticsearch, MongoDB) workloads.
• Regularly monitor resource usage and scale node pools accordingly.

Monitoring & Alerts

• Integrate resource monitoring with Prometheus/Grafana.
• Configure alerts to notify resource constraints or spikes in utilization.

---

Resource Quotas & Limit Ranges

Apply resource quotas and limit ranges to manage resource consumption within Kubernetes namespaces effectively.

Resource Quota Example

Set resource quotas at namespace level to prevent overconsumption:

apiVersion: v1
kind: ResourceQuota
metadata:
  name: prisme-quota
  namespace: prisme-ai
spec:
  hard:
    requests.cpu: "16"
    requests.memory: "32Gi"
    limits.cpu: "32"
    limits.memory: "64Gi"
    pods: "50"

LimitRange Example

Define limit ranges to enforce default limits for all pods:

apiVersion: v1
kind: LimitRange
metadata:
  name: prisme-limit-range
  namespace: prisme-ai
spec:
  limits:
    - default:
        cpu: "500m"
        memory: "1Gi"
      defaultRequest:
        cpu: "250m"
        memory: "512Mi"
      type: Container

---

Persistent Storage Recommendations

Ensure reliability and durability by properly configuring persistent volumes (PVs):

• Use Persistent Volume Claims (PVCs) with dynamically provisioned storage classes.
• Recommended storage classes: SSD-backed storage (e.g., AWS EBS GP3, Azure Premium SSD, Google Persistent Disk SSD).
• Regular backups and snapshotting via cloud provider capabilities or dedicated backup solutions.

---

Monitoring Kubernetes Resources

Use Prometheus and Grafana to continuously monitor resources:

• CPU and memory utilization dashboards.
• Persistent storage performance metrics.
• Alerting rules for resource exhaustion.

Example Prometheus alert:

groups:
- name: resource-alerts
  rules:
  - alert: HighCpuUsage
    expr: sum(rate(container_cpu_usage_seconds_total[5m])) by (pod) > 0.9
    for: 5m
    labels:
      severity: warning
    annotations:
      summary: High CPU usage detected

---

Troubleshooting Resource Issues

Common issues and resolutions:

Pod Scheduling Issues

• Verify sufficient cluster resources available:

kubectl describe pod <pod-name> -n prisme-ai

• Adjust resource requests if necessary.

High CPU or Memory Usage

• Inspect pod-level resource usage:

kubectl top pods -n prisme-ai

• Scale horizontally or vertically as needed.

Persistent Storage Issues

• Check PVC and PV status:

kubectl get pvc -n prisme-ai
kubectl describe pvc <pvc-name> -n prisme-ai

• Ensure proper storage class configuration and available storage space.

---

Next Steps

High Availability

Configure high availability setups

Prometheus & Grafana

Set up monitoring tools

Products Configuration

Configure your Prisme.ai AI products

Operations Management

Learn about scaling operations efficiently