baseline-monitoring

name	baseline-monitoring
description	Index of baseline metric families for Razorpay services. Use it to determine which standard metrics must exist for HTTP, gRPC, workers, egress, outbox, Go runtime, and relevant infra components.

Baseline Observability

Defines the standard metric families agents must consider when working on service observability.

This skill covers metrics only. Logging and tracing standards are separate and should be loaded from their own skills or references when needed.

Introduction

Observability in this skill is divided into three buckets:

• Baseline metrics: compulsory metrics every service surface must emit for the work it owns. These are the standard traffic, latency, error, health, and runtime signals that should exist whether the service is being created from scratch or improved incrementally.
• Infra metrics: metrics emitted by shared platform components such as Traefik, Kubernetes, Kafka, ALB, RDS, ElastiCache, SQS, and SNS. These are still part of the standard observability picture, but they should only be loaded when the service actually depends on or is fronted by those infra layers.
• Application metrics: service-specific business or domain metrics. The agent is free to define these as needed. They are valuable, but they do not replace the compulsory baseline metrics.

Usage Rules

• Always read references/baseline-metrics.md first.
• Then load only the references relevant to the surfaces involved in the task.
• Use the surface-to-reference map below mechanically: if the service has that surface, the linked metric families must exist.
• If the task only touches gRPC, read references/grpc-metrics.md and skip references/http-metrics.md.
• If the task only touches service code, do not load references/infra/ files unless the change explicitly involves those components.
• If the service owns workers, egress, or outbox flows, those metric families are part of the baseline for that workload and should be loaded.
• If the repo is a Go service, read references/go-runtime-metrics.md.
• Do not treat this skill as an implementation guide. It is an index of required or relevant metric families and their purposes.

Surface To Metrics Map

Use this section as the default routing rule.

If the service has this	Then make sure these metrics exist	Read
Any owned request or job-handling surface	Traffic, latency, outcome, health, and runtime baseline coverage	references/baseline-metrics.md
HTTP or REST ingress	HTTP request, status, latency, and error metric families	references/http-metrics.md
gRPC ingress	gRPC request, method, status, latency, and error metric families	references/grpc-metrics.md
Outbound or downstream calls	Egress traffic, latency, and error metric families	references/egress-metrics.md
Background workers or queue processors	Queue depth, processing rate, failure, and duration metric families	references/worker-metrics.md
Outbox flow	Outbox store, handler, backlog-age, and lifecycle-time metric families	references/outbox-metrics.md
Go runtime owned by the repo	Go process and runtime metric families	references/go-runtime-metrics.md
Traefik in front of the service	Traefik request-volume and latency metrics	references/infra/traefik-metrics.md
Edge or CDN layer relevant to the service	Edge latency, error, cache, geo, bandwidth, and certificate metrics	references/infra/edge-metrics.md
Uptime or health ownership is relevant	Uptime, health-check, and availability metrics	references/infra/uptime-metrics.md
Canary rollout analysis exists	Canary versus baseline comparison metrics	references/infra/canary-metrics.md
Kubernetes runtime is part of the service view	Pod, deployment, node, quota, and cluster-event metrics	references/infra/kubernetes-metrics.md
Kafka is part of the service topology	Kafka throughput, message-rate, and lag metrics	references/infra/kafka-metrics.md
ALB fronts the service	ALB request, latency, error, and healthy-host metrics

Reference Index

Core

• references/baseline-metrics.md: compulsory baseline metric families and how to choose the additional files that apply.
• references/http-metrics.md: HTTP and REST ingress metrics.
• references/grpc-metrics.md: gRPC ingress metrics.
• references/egress-metrics.md: outbound dependency and external API metrics.
• references/worker-metrics.md: async worker and queue-processing metrics.
• references/outbox-metrics.md: outbox lifecycle metrics.
• references/go-runtime-metrics.md: Go process and runtime health metrics.

Infra

• references/infra/traefik-metrics.md: Traefik request and latency metrics.
• references/infra/edge-metrics.md: edge, CDN, and perimeter metrics.
• references/infra/uptime-metrics.md: uptime, SLA, and health-check metrics.
• references/infra/canary-metrics.md: canary-versus-baseline comparison metrics.
• references/infra/kubernetes-metrics.md: pod, deployment, node, quota, and cluster health metrics.
• references/infra/kafka-metrics.md: Kafka throughput and lag metrics.
• references/infra/aws-alb-metrics.md: ALB request, latency, error, and healthy-host metrics.
• references/infra/aws-asg-metrics.md: ASG capacity, health, and scaling metrics.
• references/infra/aws-rds-metrics.md: RDS capacity, performance, and error metrics.
• references/infra/aws-elasticache-metrics.md: ElastiCache usage and health metrics.
• references/infra/aws-sqs-sns-metrics.md: SNS publish and SQS queue backlog metrics.

Examples

New gRPC service:

• Read references/baseline-metrics.md
• Read references/grpc-metrics.md
• Read references/go-runtime-metrics.md
• Read infra references only if the service setup includes those layers

Existing HTTP API adding a new downstream call:

• Read references/baseline-metrics.md
• Read references/http-metrics.md
• Read references/egress-metrics.md
• Read references/go-runtime-metrics.md for Go runtime coverage

Worker-only service using SQS and outbox:

• Read references/baseline-metrics.md
• Read references/worker-metrics.md
• Read references/outbox-metrics.md
• Read references/infra/aws-sqs-sns-metrics.md

Decision Rule

When working on a repo, the agent should reason in this order:

1. What surfaces does this service have?
2. For each surface, which reference file says what metric families must exist?
3. Which compulsory baseline families are currently missing or incomplete?
4. Which application metrics should be added in addition to the baseline?

Use this skill as a checklist of required metric families.

PR Requirement

At the end of the implementation, the agent must list every metric it added in the PR description.

For each metric, include:

• metric name
• short description of what it measures
• what are the normal and abnormal values for the metric, and how to interpret them
• potential alerting thresholds if applicable