Metrics

Laminar collects metrics from multiple sources and stores them in GrepTimeDB.

Collection Architecture

Desktop Mode

Vector scrapes Prometheus metrics directly from Laminar components:

Laminar Backend ──► Vector ──► GrepTimeDB
  - Controller (8004)
  - Workers (6901-6920)

Kubernetes Mode

Vector runs as a DaemonSet collecting metrics from multiple sources:

┌─────────────────────┐
│ Laminar Engine:8004 │──► Laminar-Metrics datasource
└─────────────────────┘

┌─────────────────────┐
│ Vector host_metrics │──┐
├─────────────────────┤  │
│ kubelet cAdvisor    │──┼──► GrepTimeDB-Metrics datasource
├─────────────────────┤  │
│ kube-state-metrics  │──┘
└─────────────────────┘

Metric Sources

Laminar Engine Metrics

Scraped directly from laminar-engine:8004/metrics.

Controller Metrics

Worker Metrics

GrepTimeDB Metric Views

Pre-defined views in laminar_metric_views database for easier querying:

Node Metrics

Collected by Vector's host_metrics source.

CPU

Memory

Disk

Filesystem

Network

Container Metrics (cAdvisor)

Collected from kubelet's cAdvisor endpoint.

CPU

Memory

Network

Kubernetes Object Metrics

Collected from kube-state-metrics.

Pods

Deployments

Nodes

Querying Metrics

Grafana (Prometheus datasource)

# Slot utilization percentage
(laminar_controller_registered_slots - laminar_controller_free_slots)
  / laminar_controller_registered_slots * 100
 
# Worker message throughput
rate(laminar_worker_messages_recv[5m])
 
# Container CPU usage
rate(container_cpu_usage_seconds_total{namespace="tenant-e6data"}[5m])

GrepTimeDB SQL

-- Recent slot utilization
SELECT * FROM laminar_metric_views.slot_utilization_metrics
WHERE ts > NOW() - INTERVAL '1 hour'
ORDER BY ts DESC
LIMIT 100;
 
-- Worker throughput by operator
SELECT operator_name, SUM(messages_recv) as total_recv
FROM laminar_metric_views.worker_messages_metrics
WHERE ts > NOW() - INTERVAL '1 hour'
GROUP BY operator_name;