Intelligent Workload Management (IWM) is a dynamic resource manager that runs as a lightweight control layer in Atlas. It continuously monitors your cluster’s workload and applies automated safeguards that help keep the cluster available under overload.
IWM is made up of policies that:
Monitor cluster workloads
Respond to traffic spikes, usage changes, and resource contention
Provide automated safeguards to maintain availability under load
When these policies act, your application might see new overload errors. IWM intentionally accepts a small amount of controlled failure to improve overall availability and reduce cluster-wide outages. For best results with load-shedding policies, follow the guidance in Overload Errors.
IWM Policies
IWM policies limit the blast radius of overload, so that critical workloads can continue to run even when some operations are rejected, delayed, or terminated. The following table lists the IWM policies available in Atlas.
Important
The Adaptive Operation Rate Limiting and Query Sentinel policies are only available on MongoDB 8.3+.
IWM Policy | Description | Application Symptoms | Applies to |
|---|---|---|---|
Blocks writes on the primary node when free disk space falls below defined thresholds to prevent disk exhaustion. | Write operations may receive disk-related protection errors until space is freed or capacity is increased. | Replica sets on supported versions. | |
Controls the rate at which MongoDB accepts new operations to protect the server from overload. | Under heavy load, you might see increased latency or overload errors while the cluster remains available. | Replica sets on supported versions. | |
Protects cluster availability by terminating long-running query operations. | Long-running or unbounded queries may be killed and return specific termination errors. Other workloads continue to run. | Replica sets on supported versions. |