Azure Managed Redis does not show "maxmemory" in the result of the "info" command

Mathieu Longtin 116 Reputation points Microsoft Employee
2025-10-10T20:20:06.5466667+00:00

Azure Cache for Redis and open source Redis both return the value for "maxmemory" when calling the "info" command. But Azure Managed Redis does not. It also doesn't show any of the "_perc" values that show the percentage of memory used.

Is there anyway for a client to figure what the maximum size of an "Azure Managed Redis" instance?

Azure Cache for Redis
Azure Cache for Redis
An Azure service that provides access to a secure, dedicated Redis cache, managed by Microsoft.
0 comments No comments
{count} votes

Answer accepted by question author
  1. Swapnesh Panchal 1,380 Reputation points Microsoft External Staff Moderator
    2025-10-11T00:01:19.7266667+00:00

    Hi Mathieu Longtin ,
    Thanks for contacting to Microsoft QA.
    Azure managed redis doesn’t expose maxmemory via INFO, and CONFIG is blocked. the service enforces the memory ceiling outside of redis (container/host quota), so there’s no redis-protocol call a client can use to read “maximum size.”
    so you can’t get max from INFO. get capacity from the resource’s sku (portal/arm/cli), derive total by shards, and monitor usage with azure monitor.

    way to know the max:

    • treat the management plane as source of truth. the sku you provisioned defines capacity.
    • get it from portal/arm/cli/sdk, then map sku → memory per shard from the sizing table and multiply by shard count (that’s your total data memory, before overhead/replicas).
      • classic cache: az redis show -g <rg> -n <name> --query "sku"
      • enterprise/managed variants: az redisenterprise show -g <rg> -n <name> --query "sku"
    • if you need a percentage at runtime, compute it yourself: used_memory / sku_capacity. azure managed redis doesn’t return the _perc fields.

    monitor in production:

    • rely on azure monitor metrics (e.g., used memory / rss) or INFO memory’s used_memory to track growth.
    • alert around 75–80% of sku capacity to leave headroom for fragmentation, replication, and service overhead—evictions and latency usually show up before 100%.

0 additional answers

Sort by: Most helpful

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.