Skip to main content

Monitoring

Monitor cluster health, resource utilization, and application performance through the Platform interface.

Cluster Metrics​

Overview Dashboard​

View cluster-wide metrics at a glance:

  • CPU Utilization: Total CPU usage across all nodes
  • Memory Utilization: Total memory usage across all nodes
  • Pod Count: Number of running pods
  • Node Status: Healthy vs. unhealthy nodes
  • Storage Usage: Persistent volume utilization

Node Metrics​

Monitor individual node performance:

  • CPU: Current usage, requests, limits, capacity
  • Memory: Current usage, requests, limits, capacity
  • Disk: Used vs. available disk space
  • Network: Ingress/egress bandwidth
  • Pods: Number of pods running on node
  • Conditions: Ready, MemoryPressure, DiskPressure, PIDPressure

Pod Metrics​

Track pod-level resource consumption:

  • CPU Usage: Current CPU utilization
  • Memory Usage: Current memory utilization
  • Restart Count: Number of pod restarts
  • Age: How long pod has been running
  • Status: Running, Pending, Failed, etc.

Application Performance​

Container Metrics​

Monitor individual container performance:

  • CPU Usage: Per-container CPU consumption
  • Memory Usage: Per-container memory consumption
  • Disk I/O: Read/write operations per second
  • Network I/O: Bytes sent/received
  • Resource Limits: How close to limits

Service Metrics​

Track service-level metrics:

  • Request Rate: Requests per second
  • Error Rate: Failed requests percentage
  • Latency: Response time percentiles (p50, p95, p99)
  • Active Connections: Current connection count

Alerts & Notifications​

Setting Up Alerts​

Create alerts for critical conditions:

  1. Go to Platform → Monitoring → Alerts
  2. Click Create Alert
  3. Configure alert:
    • Metric: CPU, memory, disk, etc.
    • Threshold: When to trigger alert
    • Duration: How long condition must persist
    • Severity: Critical, warning, info
  4. Set notification channels:
    • Email
    • Slack
    • PagerDuty
    • Webhook
  5. Click Create Alert

Common Alert Rules​

Pre-configured alerts for common issues:

  • High CPU Usage: Node CPU > 80% for 5 minutes
  • High Memory Usage: Node memory > 85% for 5 minutes
  • Pod Crashes: Pod restart count > 5 in 10 minutes
  • Disk Space Low: Node disk usage > 85%
  • Node Not Ready: Node becomes NotReady

Alert States​

Alerts can be in these states:

  • OK: Condition not met, no issue
  • Pending: Condition met, waiting for duration
  • Firing: Alert triggered, notifications sent
  • Resolved: Condition no longer met

Health Checks​

Liveness Probes​

Check if container is running:

  • HTTP GET: Check endpoint returns 200-399
  • TCP Socket: Check port is open
  • Exec: Run command, check exit code 0
  • Failure Action: Restart container

Readiness Probes​

Check if container ready to serve traffic:

  • Same Methods: HTTP, TCP, Exec
  • Failure Action: Remove from service endpoints
  • Use Case: Don't send traffic until ready

Startup Probes​

Check if application has started:

  • Use Case: Slow-starting applications
  • Disables Other Probes: Until startup succeeds
  • Failure Action: Restart container

Performance Dashboards​

Pre-Built Dashboards​

Access ready-made dashboards:

  • Cluster Overview: High-level cluster health
  • Node Performance: Per-node resource usage
  • Namespace Usage: Resource consumption by namespace
  • Pod Performance: Individual pod metrics
  • Storage Metrics: PV/PVC utilization

Custom Dashboards​

Create your own dashboards:

  1. Go to Monitoring → Dashboards
  2. Click Create Dashboard
  3. Add panels:
    • Select metric to visualize
    • Choose visualization type (line, bar, gauge)
    • Set time range and refresh interval
  4. Arrange panels in layout
  5. Save and share dashboard

Metrics Collection​

Prometheus Integration​

Metrics are collected via Prometheus:

  • Scrape Interval: Every 30 seconds
  • Retention: 15 days by default
  • Storage: Persistent volume for metrics
  • Query Language: PromQL for custom queries

Custom Metrics​

Expose application metrics:

# Python example with prometheus_client
from prometheus_client import Counter, Gauge, Histogram
import time

# Define metrics
requests_total = Counter('app_requests_total', 'Total requests')
active_users = Gauge('app_active_users', 'Active users')
request_duration = Histogram('app_request_duration_seconds', 'Request duration')

# Instrument code
@request_duration.time()
def handle_request():
    requests_total.inc()
    # Your application logic
    time.sleep(0.1)

Troubleshooting with Monitoring​

High CPU Usage​

  1. Check Node Metrics to identify affected node
  2. View Pod Metrics to find high-CPU pods
  3. Check pod Logs for issues
  4. Scale deployment or increase resources

Memory Leaks​

  1. Monitor Memory Usage over time
  2. Look for steadily increasing memory
  3. Check for OOMKilled pods in events
  4. Increase memory limits or fix application

Slow Response Times​

  1. Check Service Metrics for latency spikes
  2. Correlate with CPU/Memory usage
  3. Review Pod Logs for errors
  4. Check Network I/O for bottlenecks
Best Practice

Set up alerts before issues occur. Monitor trends over time to catch problems early and plan capacity.