Skip to content

SLA.md

Latest commit

 

History

History
57 lines (42 loc) · 2.87 KB

SLA.md

File metadata and controls

57 lines (42 loc) · 2.87 KB
Raw

Service Level Agreements (SLA)

This document outlines the uptime commitments and Service Level Objectives (SLOs) for the Credence Backend. It is intended for downstream integrators who rely on our APIs to build client applications, user interfaces, or third-party backend integrations.

Core Commitments

We measure and report on our service availability based on the following key Service Level Indicators (SLIs):

  • Availability (Uptime): The percentage of time the API successfully serves valid requests.
  • Latency: The time taken to process a request and return a response.

Our target Service Level Agreement (SLA) for downstream integrators is 99.9% uptime during any calendar month.

Per-Endpoint SLOs

Below are the SLO numbers for our most critical endpoints. These numbers represent our internal targets to ensure we meet the overarching 99.9% SLA.

1. Trust Score (GET /api/trust/:address)

  • SLI - Uptime: 99.95% successful responses (HTTP 200).
  • SLI - Latency: 95th percentile (P95) response time under 150ms.
  • Example Usage: 
    curl https://api.credence.example.com/api/trust/GDJ...
  • Failing Scenario: Responses taking >500ms or HTTP 5xx errors count against the error budget.

2. Bond Status (GET /api/bond/:address)

  • SLI - Uptime: 99.9% successful responses (HTTP 200).
  • SLI - Latency: P95 response time under 200ms.
  • Example Usage: 
    curl https://api.credence.example.com/api/bond/GDJ...

3. Attestations (GET /api/attestations/:address & POST /api/attestations)

  • SLI - Uptime (Reads): 99.9% successful responses.
  • SLI - Uptime (Writes): 99.5% successful responses (due to heavier validation and downstream persistence).
  • SLI - Latency (Reads): P95 response time under 150ms.
  • SLI - Latency (Writes): P95 response time under 400ms.

4. Health Checks (GET /api/health & GET /api/health/live)

  • SLI - Uptime: 99.99% availability.
  • SLI - Latency: P99 response time under 50ms.
  • Usage: Integrators should rely on /api/health/live for simple liveness probes and /api/health to verify that our dependencies (database, Redis) are functioning properly.

Monitoring and Exclusions

These SLAs apply to production environments. We actively monitor these SLIs via our Prometheus/Grafana stack (see Observability).

Exclusions: The following scenarios do not count against the SLA error budget:

  • Scheduled maintenance windows (communicated at least 48 hours in advance).
  • Rate limit rejections (HTTP 429).
  • Client-side validation errors (HTTP 400).
  • Outages caused by underlying blockchain network (e.g., Stellar Horizon) instability.

Support and Escalation

If you observe performance degrading below these thresholds for a sustained period, please verify our [status page] and open an issue in the Credence Support portal.