Anmol HandaChaos with Anmol
Home / Reels / Chaos & Reliability

Reel · Chaos & Reliability · Oct 19, 2026

Multi-AZ means you're safe?

▶ Watch on Instagram

In short

In May 2026, cooling failed in one AWS data-center hall in us-east-1, and one zone (use1-az4) was down for more than a day. Multi-AZ is a claim until you test it. The trap: zone names are different in every account, so your us-east-1a may not be their use1-az4. Take one zone away on purpose, in staging, and watch what actually breaks.

Checklist: what to do about it

  • Map your AZ names to AZ IDs (like use1-az4). Zone names differ per account; IDs don't.
  • Check that every tier (compute, database, cache, queues) really spans more than one AZ.
  • Check you have enough capacity to lose one AZ without falling over.
  • Take one zone away on purpose, in staging, and watch what actually breaks.
  • Test gray failures too: slow, not just down.
  • Write down how you'd evacuate an AZ, and who decides.

Find your AZ IDs

This lists the zone name to zone ID mapping for your account:

aws ec2 describe-availability-zones --region us-east-1 \
  --query 'AvailabilityZones[].[ZoneName,ZoneId]' --output table

Sources

Caption sources: Broadcom status notice (use1-az4), Quartz, May 2026.

Related