Azure Storage redundancy: LRS to RA-GZRS
Azure Storage redundancy is the setting that decides how many copies of your data exist and where they sit. The decision rule is to name the failure you must survive, then take the cheapest option that survives it: a datacenter outage needs ZRS, a regional outage needs GRS or GZRS, and reading data during a regional outage without waiting for a failover needs the RA- version of either.
The decision table
Every option keeps three copies in the primary region. The geo options add a second set in the paired region, which is chosen by Azure and cannot be changed.
| Option | Primary region | Secondary region | Survives | Secondary readable? |
|---|---|---|---|---|
| LRS | 3 copies in one datacenter | None | Drive, server, rack failure | n/a |
| ZRS | 3 or more availability zones | None | Loss of a datacenter (zone) | n/a |
| GRS | LRS | LRS, copied asynchronously | Regional outage (after failover) | No |
| RA-GRS | LRS | LRS, copied asynchronously | Regional outage | Yes, at the -secondary endpoint |
| GZRS | ZRS | LRS, copied asynchronously | Zone and regional outage (after failover) | No |
| RA-GZRS | ZRS | LRS, copied asynchronously | Zone and regional outage | Yes |
Three details sit under the table.
- Synchronous versus asynchronous. Writes inside the primary region are synchronous. The copy to the secondary region is asynchronous, so a regional disaster can lose the most recent writes. The gap is the recovery point objective. Without extra options there is no guaranteed lag; the opt-in geo priority replication feature adds an SLA that keeps block blob replication lag at 15 minutes or less for 99% of the billing month.
- Failover is not automatic read access. With GRS or GZRS the secondary data cannot be read or written until a failover promotes it. Only RA-GRS and RA-GZRS expose it for reading beforehand. After a failover, the secondary becomes the primary and accepts writes.
- Durability numbers. Microsoft states at least 11 nines for LRS, 12 for ZRS and 16 for GRS and GZRS over a year. Exams rarely ask for these, but they explain why the order of the options matches the order of the cost.
Redundancy is not a backup
Deletes and overwrites are applied to every copy at once. A user who deletes a container deletes it in all replicas, in all regions. Protection against that comes from soft delete, versioning, point-in-time restore and immutability, which are separate features. A question that describes accidental deletion and offers GZRS as one option is testing exactly this.
A worked example
A company has three workloads in one storage-account design. Each needs its own account, because the redundancy setting applies to the whole account and every service in it.
- Rendered thumbnails, regenerated from source images in minutes. Losing them costs a rebuild. LRS is enough and is the cheapest.
- Invoice PDFs for a web portal that must keep working if one datacenter fails, with data kept inside one region for governance reasons. ZRS fits: it survives a zone failure and has no copy outside the region.
- Legal archive that must survive the loss of the whole region, and a read-only status page must still list documents during that outage. RA-GZRS gives zone resilience in the primary region plus a readable secondary. If the status page did not need to read during the outage, GZRS would do and cost less.
Notice how the steps went: failure first, requirement words second, price last.
Which exams test this
| Exam | How the topic appears | Where covered on this site |
|---|---|---|
| AZ-900 | Architecture and services domain: describe the storage redundancy options | Architecture and services |
| AZ-104 | Storage domain: configure Azure Storage redundancy when creating and configuring storage accounts | Storage explained |
| AZ-305 | Data storage domain: recommend a data solution for protection and durability. Business continuity domain: recommend high availability for semi-structured and unstructured data | Data storage and business continuity |
How the questions are built
Questions give a failure and ask for the option, or give an option and ask what it fails to cover. The words that decide the answer:
- “Datacenter” or “zone” rules out LRS. Plain GRS also leaves the primary region on LRS, although a second copy exists elsewhere.
- “Region” or “disaster” rules out LRS and ZRS, which have no second region.
- “Read … during the outage” or “without failover” requires an RA- option. GRS and GZRS are the distractors.
- “Both” (zone failure and regional outage) points to GZRS.
- “Lowest cost” or “easily recreated data” points to LRS.
- “Data must stay in the region” points to LRS or ZRS.
The usual trap is choosing GRS because “geo” sounds safest, when the stem demands zone resilience in the primary region or reads from the secondary. A second trap is Azure Files: it supports LRS, ZRS, GRS and GZRS but not the read-access variants. Managed disks support only LRS and ZRS, and the archive blob tier is not available on ZRS, GZRS or RA-GZRS accounts.
Question 1. A storage account holds customer documents. The business requires that the data survive the loss of one datacenter in the primary region and also survive a complete regional outage. Cost should be kept as low as possible, and nobody needs to read the data from the secondary region before a failover. Which redundancy option should you choose?
- A. ZRS
- B. GRS
- C. GZRS
- D. RA-GZRS
Show answer
Answer: C
GZRS keeps three copies across availability zones in the primary region, which covers the loss of a datacenter, and replicates asynchronously to a secondary region, which covers a regional outage. GRS keeps only LRS in the primary region, so it does not give zone resilience there. ZRS has no second region. RA-GZRS also works but costs more and adds read access that the requirement does not ask for.
Question 2. A company stores audit logs in a storage account that uses RA-GRS. A technician deletes a container by mistake. Another administrator proposes reading the container back from the secondary region. What is the result?
- A. The data is still in the secondary region, because RA-GRS keeps a separate historical copy
- B. The data is deleted from the secondary region too, so recovery must come from soft delete or versioning
- C. The data can be restored by switching the account to GZRS
- D. The data can be restored only after a customer-managed failover
Show answer
Answer: B
Replication applies deletes to every copy, so the container is removed from the secondary region as well, once replication catches up. Redundancy protects against hardware and site failure, not against data-modifying operations. Recovery relies on features such as container soft delete or versioning. Read access to the secondary does not keep deleted data, and changing the redundancy option does not restore anything.
Go deeper
- Data redundancy in Azure Storage, the Microsoft Learn page the table above follows.
- Disaster recovery and storage account failover, on what a failover does and how much data it can lose.
- Azure availability zones, for what a zone physically is.
- On this site: the AZ-104 cheat sheet, the AZ-305 cheat sheet, and how to turn an outline into a plan in how to read a skills outline.