Azure Redundancy Options: LRS, ZRS, GRS, GZRS Explained

Storage redundancy is one of the most reliable topics on AZ-900, partly because it is easy to turn into a scenario question: "A team needs data to survive the loss of an availability zone but stay in one region. Which option fits?" The answer comes from keeping six acronyms straight, and the trick is that they are built from just three ideas: how many copies, where they live, and whether the second region is readable.

This guide takes the options one at a time, then puts them in a table and walks through the failures each one survives. It is part of the architecture and services area described in what is on the AZ-900 exam. Details are as documented by Microsoft in November 2026.

Two building blocks: zones and regions

Redundancy only makes sense once you know the two layers it is built on.

An availability zone is a separated group of one or more datacenters inside a single region, with independent power, cooling and networking. A region that supports zones has at least three. Zones protect against a datacenter-level failure, but not against an outage of the whole region. Not every region has zones: West US is one example of a region listed without them.

A region pair is two regions that Microsoft associates for geo-replication. Customers cannot define their own pairs. Geo-redundant storage is the typical service that replicates automatically to the paired region, and the secondary region is determined by the primary and cannot be changed.

Everything below combines those two layers. One setting applies to the whole storage account, which means to its blobs, files, queues and tables alike.

The four core options

LRS: locally redundant storage

LRS keeps three copies of your data within a single datacenter in the primary region. It has the lowest durability of the options (at least 11 nines per year) and is the lowest cost. It protects against a drive, server or rack failure, but not against a datacenter-wide disaster such as a fire or flood. It suits data you can easily recreate, or data that must stay inside one region for governance reasons.

ZRS: zone-redundant storage

ZRS copies your data synchronously across three or more availability zones in the primary region, keeping three copies with one in each zone. Durability is at least 12 nines. If one zone goes down, the data stays readable and writable. Because a write succeeds only after it is committed across the zones, writes are synchronous. ZRS does not protect against a regional disaster that affects multiple zones.

GRS: geo-redundant storage

GRS stores your data in LRS fashion in the primary region, then copies it asynchronously to the paired secondary region, hundreds of miles away, where it is again kept with LRS. That is six copies in total: three in each region. Durability is at least 16 nines. Because the geo-copy is asynchronous, a regional disaster could lose the most recent writes. With plain GRS, the secondary data is not readable until a failover occurs.

GZRS: geo-zone-redundant storage

GZRS is GRS with a better primary: ZRS in the primary region (three or more zones, synchronous), then an asynchronous copy to the secondary region, again with LRS. It also keeps six copies and has 16 nines of durability. The difference from GRS is zone redundancy in the primary region, not durability. Microsoft names GZRS as the recommended option for maximum consistency, durability, availability and disaster recovery.

The read-access variants

RA-GRS is GRS plus read access to the secondary region at all times. RA-GZRS is the same upgrade to GZRS. "RA" stands for read access. The secondary endpoint adds "-secondary" to the account name, so an app can read from it without waiting for a failover.

Read availability is where they differ in a number you may see: for the hot tier, read availability is 99.9% for GRS and GZRS, and 99.99% for RA-GRS and RA-GZRS. Write availability is 99.9% for all of them. These are Microsoft's published figures as of November 2026.

Side by side

A common wrong answer is "GRS keeps three copies." It keeps six. Another is thinking GZRS is more durable than GRS. Both are at 16 nines, and the extra protection is against a zone failure in the primary region.

What each option survives

Walk the failures from smallest to largest.

Region-wide protection comes from GRS and GZRS only, and recovery requires failing over: if the primary region is unavailable, you can fail over to the secondary, DNS is repointed and the secondary becomes the new primary.

Scenario shortcuts that map directly to questions:

• "Cheapest, and the data can be recreated" is LRS.
• "Survive the loss of a zone, keep everything in one region" is ZRS.
• "Survive an entire regional disaster" is GRS or GZRS.
• "Survive a regional disaster and a zone failure" is GZRS.
• "Read from the secondary during a primary outage with no failover" is RA-GRS or RA-GZRS, not plain GRS.

Limits and traps that show up

• Redundancy is not a backup. It protects against hardware failure, not against deletion or overwrite. All copies reflect the same current state, so an accidental delete is replicated.
• Geo-redundancy needs a pair. It requires a region that has one, and GZRS also needs availability zone support in the primary region.
• Azure Files does not support RA-GRS or RA-GZRS. Its geo-redundancy is limited to standard shares, and premium shares use LRS or ZRS only.
• Managed disks support only LRS and ZRS.
• The archive access tier is supported only on LRS, GRS and RA-GRS accounts. It is not available on ZRS, GZRS or RA-GZRS accounts.
• Deploying in a paired region does not give you failover by itself. You still design your own recovery plan.

How to learn this so it stays

Make a small grid with the six options down the side and four columns: copies, where, nines, readable secondary. Fill it from memory, check it, and repeat the next day and a week later. That is the pattern in what spaced repetition actually does, and the layout is the method from how to memorize a list, a table or a formula sheet. The rest of the plan for the exam is in how to study for AZ-900.

Redundancy often sits next to the governance tools in scenario questions, because both decide what survives what. If you want the other half of that comparison, read Azure RBAC vs Policy vs resource locks. To test yourself on these options, AZ-900 practice questions includes redundancy and availability zone cards.

The cards behind this post are in the Azure Fundamentals course, in the Storage Services unit. The course is unofficial and is not affiliated with or endorsed by Microsoft. Cards are recall practice, not a substitute for opening a storage account in the portal and looking at its redundancy setting yourself.