Skip to content
AutoObserve
Design Partner

EVIDENCE STORAGE

ClickHouse Preserve evidence for extended investigation windows.

Production investigations often need history beyond live telemetry retention. ClickHouse stores historical evidence — metrics, logs and traces — so AutoObserve can reason over longer incident windows.

Evidence
Historical evidence
Role
Evidence storage layer
Enables
Long-term investigation
Status
Preview
See the evidence received ↓View architecture →

Read integration docs →

Telemetry

  1. ClickHouse

  2. Historical evidence

  3. Investigation

  4. Decision

WHY THIS INTEGRATION MATTERS

Investigations need history — not just live signals.

When checkout-api degrades intermittently, live retention may miss the earlier pattern. ClickHouse preserves evidence so investigations span days, not minutes.

  • Extended retention

    Query evidence beyond default live telemetry windows.

  • Historical patterns

    Compare current degradation to prior incident windows.

  • Long-window correlation

    Join historical metrics, logs and change events in one query plan.

  • Post-incident review

    Preserve evidence for RCA and design-partner validation.

EVIDENCE RECEIVED

Historical evidence in ClickHouse

Normalized telemetry and change events persist as queryable historical evidence — available during extended investigations.

  • Historical evidence

  1. Ingest

    Write normalized evidence to ClickHouse tables.

  2. Retain

    Apply retention policies aligned with investigation needs.

  3. Query

    Retrieve historical windows during RCA and post-incident review.

  4. Investigate

    Use long-window evidence inside connected investigations.

HOW AUTOOBSERVE USES THIS EVIDENCE

Evidence becomes queryable investigation history.

ClickHouse is not a replacement observability UI — it is the storage layer that preserves evidence AutoObserve connects and investigates over time.

Provides

  1. Historical evidence

AutoObserve creates

  1. Long-term investigation
  2. Historical correlation
  3. Post-incident review
  4. Extended retention

ClickHouse is memory. AutoObserve is understanding over time.

Read: The evidence model →

EVIDENCE STORAGE ARCHITECTURE

Persist connected evidence for long-window investigations.

Ingest, tier and query paths keep historical evidence available without duplicating live telemetry pipelines.

Write normalized metrics, logs and traces to ClickHouse.

  1. Live telemetry

  2. Evidence writer

  3. ClickHouse

QUICK START

Connect ClickHouse evidence storage

Configure the storage layer and retention policy, then verify historical checkout-api evidence is queryable.

  1. Deploy ClickHouse

    Provision ClickHouse for historical evidence storage in your environment.

    Integrations reference →

  2. Configure retention

    Set retention windows aligned with investigation and compliance needs.

    Architecture overview →

  3. Verify historical queries

    Confirm prior checkout-api incidents are queryable beyond live retention.

    Investigate evidence →

WHAT SUCCESS LOOKS LIKE

History preserved. Long-window investigation ready.

Success means historical evidence is queryable during RCA — not merely that ClickHouse is running.

  • Retention configured

    Evidence persists beyond default live telemetry window.

  • Historical queries work

    Prior checkout-api windows return connected evidence.

  • Long-window correlation

    Historical metrics align with change events across days.

REAL INCIDENT

Same checkout spike — historical log in focus.

The canonical checkout-api degradation with log evidence highlighted — the historical signal that reveals an earlier pattern before live retention expired.

  1. Metric

    14:31:00

    Checkout latency +184%

Walk through the evidence chain · 1/6

Evidence step 1 of 6

TROUBLESHOOTING

Common ClickHouse storage failures

  • Problem

    No historical evidence stored

    Likely cause

    Writer pipeline not configured or schema mismatch on ingest.

    Resolution

    Verify evidence writer connectivity and table schema for each signal type.

    Verify

    checkout-api metrics queryable beyond live retention window.

  • Problem

    Historical window too narrow

    Likely cause

    TTL or partition policy shorter than investigation needs.

    Resolution

    Extend retention for production evidence tables.

    Verify

    Prior incident window still queryable after 7 days.

  • Problem

    Historical queries too slow for RCA

    Likely cause

    Missing indexes or unpartitioned high-cardinality tables.

    Resolution

    Review partition keys and materialized views for investigation queries.

    Verify

    Historical checkout-api query completes within investigation SLA.

PRODUCTION INTELLIGENCE

How ClickHouse contributes to production understanding

Historical storage enables long-window investigation inside connected production intelligence.

EvidenceEnables
Historical metricsLong-term performance analysis
Historical logsExtended RCA windows
Historical tracesPrior request-flow comparison

DESIGN PARTNERS

Validate this evidence source with your production stack.

Design partners connect real evidence sources and validate investigation workflows before GA — with direct engineering access.

Become a Design PartnerRead integration docsBecome a Design Partner