Prometheus Integration

Connect ReliaPulse to your Prometheus server to monitor metrics and create alerts based on PromQL queries.

Prerequisites

Prometheus server accessible from ReliaPulse
API endpoint enabled on Prometheus
Network connectivity between services

Configuration

Create Integration

Go to Dashboard → Settings → Integrations
Click Add Integration
Select Prometheus
Fill in the configuration:

Field	Description	Example
Name	Friendly name	`Production Prometheus`
URL	Prometheus API URL	`https://prometheus.example.com`
Username	Basic auth username (optional)	`admin`
Password	Basic auth password (optional)	`***`

Test Connection

Click Test Connection to verify ReliaPulse can reach your Prometheus server.

Creating Metrics Queries

Basic Query

Go to your integration settings
Click Metrics tab
Click Add Query
Enter your PromQL query

Example - CPU Usage:

avg(rate(node_cpu_seconds_total{mode!="idle"}[5m])) * 100

Example - Memory Usage:

(1 - (node_memory_MemAvailable_bytes / node_memory_MemTotal_bytes)) * 100

Example - HTTP Request Rate:

sum(rate(http_requests_total[5m]))

Query Configuration

Field	Description
Name	Display name for the metric
Query	PromQL expression
Polling Interval	How often to fetch (30s - 1h)
Unit	Display unit (%, ms, requests, etc.)
Warning Threshold	Value to trigger warning
Critical Threshold	Value to trigger critical alert

Multi-Series Metrics

For queries that return multiple time series, enable Multi-Series Mode:

# Returns one series per instance
node_cpu_seconds_total{mode="idle"}

Configure:

Group By Tags: instance, job
Aggregation: AVG, SUM, MAX, MIN
Max Series: Limit number of series tracked

Threshold Configuration

Threshold Operators

Operator	Description
Greater Than	Alert when value > threshold
Less Than	Alert when value < threshold
Equal To	Alert when value = threshold
Not Equal To	Alert when value ≠ threshold

Example Thresholds

CPU Usage (higher is worse):

Warning: 70 (>70%)
Critical: 90 (>90%)
Operator: Greater Than

Available Memory (lower is worse):

Warning: 1073741824 (<1GB)
Critical: 536870912 (<512MB)
Operator: Less Than

Linking to Components

Connect metrics to components for automatic status updates:

Edit a component
Set type to METRIC
Select your Prometheus integration
Choose or create a metrics query
Configure thresholds

When metric exceeds critical threshold:

Component status changes to MAJOR_OUTAGE
Incident created automatically (if enabled)
Notifications sent to subscribers

PromQL Tips

Rate vs Instant

Use rate() for counter metrics:

# Good - shows requests per second
rate(http_requests_total[5m])

# Bad - shows total count (always increasing)
http_requests_total

Aggregation

Aggregate across instances:

# Average across all nodes
avg(node_load1)

# Sum of all requests
sum(rate(http_requests_total[5m]))

# Max latency from any pod
max(http_request_duration_seconds)

Label Filtering

Filter by labels:

# Specific service
http_requests_total{service="api"}

# Exclude status codes
http_requests_total{status!~"5.."}

# Multiple conditions
http_requests_total{service="api", method="POST"}

Time Ranges

Common time ranges:

[1m] - Last minute
[5m] - Last 5 minutes (recommended for rate)
[1h] - Last hour
[1d] - Last day

Troubleshooting

Connection Failed

Verify Prometheus URL is correct
Check network connectivity
Verify authentication credentials
Ensure Prometheus API is enabled

No Data Returned

Test query in Prometheus UI first
Check time range of data
Verify metric name spelling
Check label values

SSRF Blocked

ReliaPulse blocks requests to private networks for security. Ensure your Prometheus server is accessible on a public IP or configure appropriate network access.

Security

Network Security

Use HTTPS for Prometheus connections
Consider VPN or private network access
Limit Prometheus API to read-only queries

Authentication

Prometheus supports:

Basic authentication
Bearer token (via reverse proxy)
mTLS (via reverse proxy)

Datadog Grafana