A blue cloud-computing network connects document databases with laptops and desktop clients.
Azure DocumentDB's "query from any node" capability is now generally available. It lets a client send a query to any node in a multi-shard cluster, instead of funneling every request through one entry point. This matters for anyone running MongoDB-compatible workloads on Azure, though the feature has a clear boundary that is easy to miss.

A blue cloud-computing network connects document databases with laptops and desktop clients. What Microsoft announced​

Abinav Rameesh, Principal PM Manager for Azure DocumentDB, announced the feature on November 17, 2026, in a post covering Ignite 2026 updates. Microsoft's limits documentation says the feature requires a multishard cluster and has no effect on a single-shard cluster.

The announcement says that in multi-shard clusters, queries can be submitted to any node. Each node can handle a request locally or route it to the appropriate shards. When a query spans shards, that node gathers the results. Microsoft's stated benefit is that routing responsibility is spread across the cluster, which simplifies access to horizontally scaled workloads.

The Azure Updates entry is labeled "Launched," which Microsoft defines as fully released and production-ready for all Azure customers. The updates page did not render the individual entry when it was retrieved. The Microsoft engineering blog post is therefore the source for the announcement wording.

How it fits with existing scaling​

Azure DocumentDB is Microsoft's managed, MongoDB-compatible database. It is built on the open-source DocumentDB project, which the Linux Foundation governs. It was previously named vCore-based Azure Cosmos DB for MongoDB, according to Microsoft's FAQ.

Microsoft's scalability documentation describes how horizontal scaling already worked. The service automatically maps logical shards to physical shards, and requests are automatically routed to the physical shard that owns the hash range for the data being queried. The documentation now adds the new behavior: with query from any node, every physical shard accepts client connections, so connection handling and query routing also scale out as shards are added.

In practice, the change is about where clients can enter the cluster. Data placement and automatic rebalancing are separate, existing service behaviors.

What it does not mean​

Microsoft's materials don't support a few readings of this feature:

  • Not a full copy of data on every node. A node that receives a query may route it to other shards and gather the results. It doesn't hold everything itself.
  • Not guaranteed single-node execution. Cross-shard queries still touch multiple shards.
  • Not a published performance claim. The sources give no throughput, latency or cost figures. Any speedup depends on your workload, shard key and query patterns.
  • Not a change to single-shard clusters. They gain nothing from it.

Who can use it​

The multi-shard requirement rules out some deployments. The documented maximum is 10 physical shards. Microsoft's limits page also says the M10, M20 and M25 tiers support only one physical shard (node). Clusters on those tiers can't meet the multi-shard prerequisite.

Other documented limits still apply to queries that cross shards:

  • In sharded clusters, if a query pulls data across nodes, the limit on that data size is 1 GB.
  • The default query execution limit is 120 seconds. You can override it per query with maxTimeMS in your MongoDB driver.

These are existing service limits, not new ones introduced with this launch.

Practical guidance for admins and developers​

Microsoft's announcement doesn't give rollout steps, connection-string changes, driver requirements or client-side load-balancing settings for this feature, so I won't invent any. Here is what the documentation supports:

  1. Check your topology. Confirm your cluster has more than one physical shard. Single-shard clusters see no effect.
  2. Check your tier. M10, M20 and M25 clusters are single-shard only.
  3. Keep the shard key in mind. Microsoft's sharding guidance says to choose the most common query filter as the shard key. That localizes searches to a single physical shard and optimizes the most transactions. Entry-point flexibility doesn't replace good shard-key design.
  4. Test your own workload. There are no published benchmarks, so measure before and after.
  5. Use standard connectivity. Clients connect with standard MongoDB drivers and the cluster's connection string, over the public endpoint with firewall rules or through Private Link.

Context from the Ignite announcement​

The same post lists other DocumentDB updates:

  • The DocumentDB extension for Visual Studio Code, now generally available.
  • Near-instant clusters on the M10 through M30 tiers, generally available.
  • Full-text search, in public preview.
  • Fine-grained role-based access control for databases and collections, in public preview.
  • DocumentDB enabled by Azure Arc, in public preview.
  • DocumentDB on Azure Local, in public preview.

Query from any node is the only scaling item in the list that has reached general availability.

Analysis​

This is a modest but sensible production-readiness step. In many sharded systems, the entry point can become a bottleneck or a design headache. Letting every shard accept connections removes one scaling constraint as clusters grow. The catch is that the benefit applies only to multi-shard clusters, and Microsoft hasn't published numbers. Treat it as an architectural option to evaluate, not a guaranteed speed boost.

 

References

  1. (Launched) Generally Available: Azure DocumentDB elasticity—query from any node Azure Updates 2026-11-17T20:00:32Z
  2. Azure DocumentDB at Microsoft Ignite 2026: What's New devblogs.microsoft.com
  3. Service limits and Quotas Reference - Azure DocumentDB | Microsoft Learn learn.microsoft.com