QCon San Francisco 2026 hosted engineering teams from fintech, e-commerce, and platform companies sharing lessons from UUID v7 migrations and adoption. The talks were practitioner-focused — real migration timelines, gotchas, and benchmark numbers from production databases. See UUID v7 Performance Benchmarks for the numbers context.

Session: Migration from UUID v4 to v7 at scale

An e-commerce platform shared their migration of a 150M-row orders table from UUID v4 to UUID v7 primary keys:

Timeline:

Outcome: Insert throughput increased 3.1× on the orders table. WHERE id > $cursor ORDER BY id pagination replaced LIMIT/OFFSET with created_at index.

See Zero-Downtime UUID Migration for the full step-by-step.

Session: UUID v7 as distributed trace IDs

A fintech platform described replacing their custom snowflake IDs with UUID v7 for trace correlation:

import { v7 as uuidv7 } from 'uuid';

// W3C traceparent: 00-<32 hex traceId>-<16 hex spanId>-01
function newTraceparent(): string {
  const traceId = uuidv7().replace(/-/g, '');         // 32 hex chars
  const spanId  = uuidv7().replace(/-/g, '').slice(0, 16); // 16 hex chars
  return `00-${traceId}-${spanId}-01`;
}

// Extract trace start time from traceId
function traceStartMs(traceparent: string): number {
  const traceId = traceparent.split('-')[1];
  return parseInt(traceId.slice(0, 12), 16);
}

UUID v7 trace IDs enabled log queries like “all traces from the last 5 minutes” without Jaeger/Tempo — just a timestamp prefix filter on the trace ID column. See UUID v7 in Microservices.

Session: PostgreSQL 18 adoption panel

A panel of five teams shared their PostgreSQL 18 upgrade timelines specifically for uuidv7():

-- Before PG 18: application-layer or pg_uuidv7 extension
CREATE TABLE orders (
  id uuid PRIMARY KEY DEFAULT gen_random_uuid()  -- UUID v4
);

-- After PG 18 upgrade: single-line change
ALTER TABLE orders
  ALTER COLUMN id SET DEFAULT uuidv7();

-- New tables: fully server-side
CREATE TABLE events (
  id uuid PRIMARY KEY DEFAULT uuidv7(),
  payload jsonb NOT NULL
);

All five teams reported the upgrade was “a single ALTER statement” for existing tables that already had UUID primary keys. See PostgreSQL 18 Native UUID v7.

Further reading

External references

Frequently asked questions

What production UUID v7 migration patterns were shared at QCon SF 2026?

Engineering teams shared dual-column migration patterns: adding an id_v7 uuid column, backfilling with uuidv7() in batches of 5,000–10,000 rows, building a CREATE INDEX CONCURRENTLY, then swapping the primary key in a brief transaction. Teams reported the PK swap taking under 500ms on tables up to 200M rows. The most common pitfall was forgetting to update foreign key columns in child tables before the cutover. See Zero-Downtime UUID Migration.

How did QCon speakers integrate UUID v7 with OpenTelemetry?

Speakers recommended stripping hyphens from a UUID v7 (uuidv7().replace(/-/g, '')) to produce a valid 32-character W3C trace ID. The first 12 hex characters are the millisecond timestamp — useful for log-range filtering without a separate start time field. UUID v7 span IDs (first 16 hex chars) sort chronologically within a trace. See UUID v7 in Microservices.