Monitoring and Alerting

Convox provides integrated metrics collection and alerting capabilities that give you comprehensive visibility into your applications and infrastructure. Built directly into the Convox platform, monitoring requires no additional tools or complex configuration.

Monitoring Capabilities

Convox Monitoring and Alerting offers:

  • Automatic metrics collection from your racks and applications
  • Pre-configured dashboards with essential infrastructure metrics
  • Custom panel creation using Smart Queries or PromQL
  • Intelligent alerting with configurable thresholds and notifications
  • Seamless integration with your existing Convox workflow

Enabling Metrics Collection

Metrics collection is enabled at the rack level through the Rack Settings page.

Steps to Enable

  1. Navigate to your rack in the Convox Console
  2. Click Rack Settings in the left sidebar
  3. Scroll to the Dashboard Settings section
  4. Toggle the Enable Metrics Agent switch to on
  5. In the plan picker modal, choose Free (lightweight Prometheus chart, no metered billing) or Paid (kube-prometheus-stack with metered billing; requires payment method on file).
  6. Wait approximately 5-15 minutes for the chart to install. Status updates in the same panel.

Once enabled, the Metrics Agent installs into your rack and begins collecting performance data from all running services.

Switching Plans

When monitoring is enabled, the Rack Settings → Dashboard Settings panel shows a Switch Plan button alongside the existing toggle. Switching between Free and Paid:

  • Uninstalls the current plan's chart, then installs the new plan's chart. Total ~5-15 minutes.
  • During the switch, convox ps GPU enrichment fields show em-dash sentinels until the new chart is installed and prometheus_url is wired (see below).
  • Free → Paid requires a payment method on file. The Console surfaces a card-on-file gate before the switch begins.
  • Paid → Free preserves the underlying Stripe subscription (metering stops; no auto-cancellation).

Setting prometheus_url for convox ps GPU enrichment

Post-3.24.6, the rack does not auto-resolve a Prometheus URL. To populate GPU fields in convox ps after enabling monitoring, set prometheus_url on your rack.

Note: the free-plan Prometheus chart depends on the Convox Console monitoring-redesign deploy. If your Convox Console version does not yet show the Free vs Paid plan picker in Rack Settings → Dashboard Settings, the free-plan service URL below will not resolve (no chart deployed). Use the paid-plan URL or wait until the redesign rolls out to your Convox Console.

# Paid plan in-cluster Prometheus:
convox rack params set prometheus_url=http://convox-kube-prometheus-sta-prometheus.convox-monitoring.svc.cluster.local:9090

# Free plan in-cluster Prometheus:
convox rack params set prometheus_url=http://prometheus-gpu-metrics-server.kube-system.svc.cluster.local:80

Until prometheus_url is set, convox ps GPU fields render as em-dash sentinels (—) even after monitoring is enabled in the Console.

Default Dashboards

When you enable metrics, Convox automatically creates default panels that provide immediate visibility into your infrastructure:

  • Rack CPU Usage - Overall CPU utilization across your rack
  • Rack Memory Usage - Memory consumption patterns
  • Rack Network I/O - Network throughput and activity
  • Rack Network Errors - Network error rates
  • Application Pod Count - Number of running application instances
  • Application Pod Ready % - Percentage of healthy application pods

These panels give you a comprehensive baseline view of your infrastructure health without any additional configuration.

Custom Panels

Create custom panels to track specific metrics that matter to your applications.

Creating a Panel

  1. Navigate to the V3 Dashboard in the Metrics section
  2. Click Create Panel
  3. Configure your panel:
    • Name: Descriptive name for your panel
    • Description: Optional details about what the panel tracks
    • Chart Type: Visualization type (automatically set for Smart Queries)

Adding Queries

Panels can contain multiple queries to compare different metrics or services.

Smart Queries

Smart Queries are pre-configured queries that make metric selection intuitive:

  1. Select your Rack from the dropdown
  2. Choose All Apps or select a specific application
  3. Choose All Services or select specific services
  4. Select a metric from the Smart Query Metrics list
  5. Click Generate Preview to verify the query

Available Smart Query metrics include:

  • CPU Usage Trend
  • Memory Usage Trend
  • Network I/O patterns
  • Request rates and response times

Custom Queries

For advanced monitoring needs, you can input any valid PromQL query to access the full range of collected metrics.

Panel Management

  • Generate Preview: View your panel before saving
  • Create Panel: Save the panel to your dashboard
  • Edit/Delete: Modify existing panels using the panel controls

Alerting

Set up intelligent alerts to be notified when your applications exceed defined thresholds.

Creating Alert Rules

  1. Navigate to Alert Manager in the Metrics section
  2. Click the settings icon to create a new alert
  3. Configure your alert:

Basic Configuration

  • Rule Name: Descriptive name for the alert
  • Severity: Choose from Warning, Critical, or other severity levels
  • Panel: Select the panel containing the metric to monitor
  • Query: Choose the specific query within the panel

Conditions

  • Condition: Set when the alert should trigger (greater than, less than, etc.)
  • Threshold: The value that triggers the alert
  • For Seconds: How long the condition must persist before triggering
  • Keep Firing For Seconds: How long to keep the alert active once triggered

Documentation

  • Summary: Brief description used in alert notifications
  • Description: Detailed explanation of what the alert monitors
  • Labels: Key-value pairs for organizing and routing alerts

Alert Management

The Alert Manager dashboard shows:

  • Active Alerts: Currently firing alerts
  • Total Alerts: All configured alert rules
  • Monitoring Status: Overall system health
  • Alert History: List of all alert rules with their current state

Notifications

Configure notification integrations to receive alerts where your team works.

Supported Integrations

  • Slack: Send alerts directly to Slack channels
  • Discord: Deliver notifications to Discord channels
  • PagerDuty: Route alerts to PagerDuty for on-call workflows

Setting Up Notifications

  1. Navigate to the Integrations page
  2. Select your preferred notification service
  3. Follow the integration-specific setup process
  4. Test the integration to ensure alerts are delivered properly

Best Practices

Panel Organization

  • Create panels that group related metrics for easier analysis
  • Use descriptive names and descriptions for panels and queries
  • Combine multiple services in a single panel to compare performance

Alert Configuration

  • Set appropriate thresholds based on your application's normal operating ranges
  • Use "For Seconds" to avoid alert noise from temporary spikes
  • Include clear summaries and descriptions for faster incident response
  • Apply consistent labeling for better alert organization

Monitoring Strategy

  • Start with default panels to understand baseline performance
  • Create custom panels for application-specific metrics
  • Set up alerts for critical thresholds that require immediate attention
  • Review and adjust alert thresholds based on historical data

Troubleshooting

Metrics Not Appearing

  • Verify the Metrics Agent is enabled in Rack Settings
  • Wait 2-3 minutes for agents to fully initialize
  • Check that your applications are running and generating metrics

Alert Issues

  • Ensure alert conditions and thresholds are correctly configured
  • Verify notification integrations are properly set up
  • Check alert timing settings (For Seconds, Keep Firing For Seconds)

Performance Impact

Convox monitoring is designed to have minimal impact on your applications. The Metrics Agent collects data efficiently without affecting application performance.

Command Line Interface

While monitoring is primarily managed through the Convox Console, you can view application logs using the CLI:

# View application logs
convox logs -a myapp

# Filter logs by service
convox logs -a myapp --service web

# View logs for a specific time period
convox logs -a myapp --since 1h

# Stream application logs (default behavior)
convox logs -a myapp

Known Limitations

  • Rack-only users (no Convox Console connection) do not have access to GPU-observability Prometheus charts. The DCGM exporter still installs via gpu_observability_enable=true, but no Prometheus chart scrapes it. Connect to Convox Console and enable monitoring to restore full functionality.
  • monitoring_metrics_provisioned has been replaced. Monitoring is now configured in the Convox Console (Rack Settings → Dashboard Settings → Enable Metrics Agent) rather than via rack parameters.
  • Chart-version overrides require a Disable→Enable cycle to take effect. Setting prometheus_gpu_metrics_chart_version or prometheus_gpu_metrics_retention on a rack with monitoring enabled does not immediately re-deploy — the new value applies on the next Disable→Enable cycle from the Console.
  • convox ps GPU fields require prometheus_url to be set. Even with monitoring enabled in the Console, the rack does not auto-resolve a Prometheus URL. See "Setting prometheus_url" above.
  • DCGM without monitoring: setting gpu_observability_enable=true on the rack without enabling monitoring in the Console results in DCGM running with no scraper. Dashboards remain empty until monitoring is enabled in the Console.
  • Prometheus-backed autoscaling requires explicit prometheus_url. Users with scale.autoscale.gpuUtilization or scale.autoscale.queueDepth triggers (without an explicit per-trigger prometheusUrl) must set prometheus_url. CPU-, memory-, and scale.keda.triggers-based autoscale (e.g. aws-sqs-queue, kafka, cron) continue to work without a Prometheus URL.
  • Plan-switch requires rack version 3.24.6+. Users on earlier rack versions cannot switch between free and paid plans via the Console. Upgrade to 3.24.6+ to use plan-switch.
  • Downgrading to 3.24.5 with monitoring enabled requires preparation. Console-installed prometheus-gpu-metrics chart in kube-system conflicts with 3.24.5's rack-managed chart. Before downgrading: disable monitoring in the Console (Rack Settings → Dashboard Settings → toggle off), wait ~5 minutes for chart uninstall to complete, then run convox rack update --version 3.24.5.
  • Free-plan Prometheus memory budget is typically <500Mi. The free chart includes the upstream default scrape jobs (kubernetes-pods, kubernetes-service-endpoints, kubernetes-nodes) in addition to the DCGM scrape. Larger clusters may need to reduce retention via prometheus_gpu_metrics_retention.

See Also

Convox Logo
Find
HomeProductPricingEnterpriseBlogDocumentation
Follow
LinkedInXGitHub
Chat
sales@convox.com
© Convox 2024 / A Curious Company
Privacy PolicyTerms of Use