Capacity Pool metrics
Momento Cache provides two limited-preview observability contracts for Capacity Pools:
- A Prometheus endpoint that you scrape. It returns seven metrics.
- Amazon CloudWatch delivery that emits six metrics when delivery is enabled.
The paths use different names, units, cadence, and eviction semantics. CloudWatch does not include the Prometheus memory-utilization ratio.
Prometheus
Send GET /capacity_pool/metrics to the regional control-plane endpoint with the same
Authorization header used for Capacity Pool reads. The response uses Prometheus text exposition
format and includes Pools for the authenticated account in that serving region.
Every sample has pool_name, account_id, and endpoint labels.
| Metric | Type and unit | Meaning |
|---|---|---|
capacity_pool_memory_used_bytes | Gauge, bytes | Valkey memory in use across primary shards. |
capacity_pool_memory_maxmemory_bytes | Gauge, bytes | Valkey maxmemory across primary shards. |
capacity_pool_memory_utilization_ratio | Gauge, 0.0–1.0 ratio | Memory used divided by maxmemory. |
capacity_pool_evictions_total | Counter, cumulative keys | Total keys evicted across primary shards. |
capacity_pool_host_cpu_usage_ratio | Gauge, 0.0–1.0 ratio | Host CPU utilization across primary and replica nodes. |
capacity_pool_host_network_rx_utilization_ratio | Gauge, 0.0–1.0 ratio | Host network receive utilization across primary and replica nodes. |
capacity_pool_host_network_tx_utilization_ratio | Gauge, 0.0–1.0 ratio | Host network transmit utilization across primary and replica nodes. |
The first four samples are emitted for each Pool with a complete primary-memory observation. A host CPU or network sample is omitted when its required reading is unavailable. You choose the Prometheus scrape cadence.
See the Capacity Pool API reference for the HTTP response and error contract.
Amazon CloudWatch
When CloudWatch delivery is enabled, Momento emits Pool data in 60-second reporting intervals
under the namespace Momento/Cache. Every datum has exactly these dimensions:
PoolAccountIdEndpoint
| Metric | CloudWatch unit | Meaning |
|---|---|---|
PoolMemoryUsedBytes | Bytes | Sum of Valkey memory in use on primary shards. |
PoolMemoryProvisionedBytes | Bytes | Sum of Valkey maxmemory on primary shards. This is allocated Pool capacity, not replica-inclusive Flex usage. |
PoolEvictions | Count | Keys evicted during the reporting interval, calculated from primary-shard counter deltas. |
PoolCpuUtilization | Percent | Host CPU utilization across primary and replica nodes. |
PoolNetworkReceiveUtilization | Percent | Receive rate relative to aggregate baseline bandwidth across primary and replica nodes. |
PoolNetworkTransmitUtilization | Percent | Transmit rate relative to aggregate baseline bandwidth across primary and replica nodes. |
Memory and utilization measurements are aggregated as statistic sets for each interval.
PoolEvictions is the sum of interval deltas. A missing CPU or network reading omits that metric
instead of reporting zero.
This section defines the emitted metric contract only. It does not define a CloudWatch enablement or IAM setup workflow.