Prometheus Integration

AtlasAI integrates with Prometheus to ingest metrics and alert data. You can use remote write to stream metrics to AtlasAI, or configure AtlasAI to scrape your Prometheus server directly.

Setup Methods

Method 1: Remote Write (Recommended)

Add a remote write target to your Prometheus configuration to stream metrics directly to AtlasAI.

Step 1: Navigate to Settings → Integrations → Prometheus in AtlasAI and copy your remote write URL and credentials.

Step 2: Add the remote write block to your prometheus.yml:

remote_write:
  - url: "https://<tenant-plane-url>:8443/api/v1/prom/write"
    headers:
      Authorization: "Bearer <INTEGRATION_API_KEY>"
    queue_config:
      max_samples_per_send: 1000
      batch_send_deadline: 5s
    write_relabel_configs:
      - source_labels: [__name__]
        regex: "up|node_.*|container_.*|kube_.*"
        action: keep

Step 3: Reload Prometheus:

kill -HUP $(pidof prometheus)

Step 4: Verify data flow in AtlasAI under Settings → Integrations → Prometheus — the status should show “Receiving data” within 60 seconds.

Method 2: Federation / Scrape

Configure AtlasAI to scrape your Prometheus server’s federation endpoint.

Step 1: Navigate to Settings → Integrations → Prometheus → Add Source

Step 2: Enter your Prometheus server details:

• URL: http://prometheus.monitoring.svc:9090
• Scrape Interval: 30s
• Metric Selectors: {__name__=~"node_.*|kube_.*"}

Step 3: Click Test Connection and then Save.

Importing Alertmanager Rules

AtlasAI can import your existing Prometheus alerting rules:

1. Navigate to Settings → Integrations → Prometheus → Alert Rules
2. Click Import Rules
3. Paste your alerting rules YAML or upload the file
4. AtlasAI converts them to native alert definitions while preserving the original PromQL expressions

PromQL Support

AtlasAI supports PromQL queries natively in the Dashboards and Logs modules. Any dashboard widget can use PromQL to query Prometheus metrics stored in AtlasAI:

rate(http_requests_total{status=~"5.."}[5m])
  / rate(http_requests_total[5m])

Troubleshooting