Sidecar memory usage can grow continuously due to configuration bloat, request buffering, metric cardinality, or HTTP/2 flow-control defaults. This topic explains each cause and how to resolve it.
Symptom
Pod-level memory monitoring shows that sidecar (Envoy) memory usage keeps increasing over time.
Causes
Cause | Description |
Configurations stored in sidecars | By default, each sidecar stores the configurations of all services in the cluster, including endpoint information. As more services are deployed, configuration volume and memory usage increase. |
Real-time request traffic | Envoy buffers incoming requests, outgoing responses, and queued requests. Under heavy traffic, these buffers consume significant memory. |
High-cardinality metrics | Metrics with a large number of unique label combinations occupy substantial memory. |
gRPC throttling-related data | The HTTP/2 codec uses |
Solutions
Reduce configuration scope
By default, sidecars receive configurations for every service in the cluster. Narrowing the scope to only the services a workload communicates with reduces memory usage.
Log on to the ASM console.
In the left-side navigation pane, choose Service Mesh > Mesh Management.
Find the target ASM instance and click Manage in the Actions column.
In the left-side navigation pane, choose Mesh Optimization Center > Sidecar recommendation.
Configure sidecar recommendation to limit which service configurations are pushed to each sidecar.
For more information, see Configuration push optimization overview.
Bypass unnecessary traffic interception
Excluding traffic that does not need mesh-level observability or policy enforcement reduces the amount of data Envoy buffers. For example, exclude connections to middleware such as Redis (port 2379) or MySQL (port 3306), or traffic originating from outside the cluster.
Log on to the ASM console. In the left-side navigation pane, choose Service Mesh > Mesh Management.
On the Mesh Management page, click the name of the ASM instance. In the left-side navigation pane, choose Dataplane Component Management > Sidecar Proxy Setting.
On the Sidecar Proxy Setting page, click the global tab. Then, click enable/disable Sidecar proxy by port or address and configure the block scope.
Reduce metric cardinality
Metrics with a large number of unique label values are a common source of memory pressure. Review custom metrics and remove or aggregate labels that produce excessive cardinality.
For more information, see Create custom metrics in ASM.
Tune HTTP/2 flow-control window sizes
The HTTP/2 codec in Envoy uses two flow-control parameters:
| Parameter | Default | Description |
|---|---|---|
initial_stream_window_size | 256 MB | Maximum bytes buffered per stream |
initial_connection_window_size | 256 MB | Maximum bytes buffered across all streams on a single connection |
When gRPC or HTTP/2 upstream and downstream services consume data slower than it arrives, Envoy accumulates data up to these limits. Lowering these values reduces the maximum memory that flow control can consume per sidecar.
Log on to the ASM console. In the left-side navigation pane, choose Service Mesh > Mesh Management.
On the Mesh Management page, click the name of the ASM instance. In the left-side navigation pane, choose Plugin Extension Center > Market Place.
On the Market Place page, click Set the initial window size for http2/grpc and configure
initial_stream_window_sizeandinitial_connection_window_sizeto values appropriate for your workload.