Metrics API¶
NOTE: The metrics API may change in the future, this serves as a snapshot of the current metrics.
Admin¶
Administrators can monitor Serving control plane based on the metrics exposed by each Serving component. Metrics are listed next.
Activator¶
The following metrics allow the user to understand how application responds when traffic goes through the activator eg. scaling from zero. For example high request latency means that requests are taken too much time be fulfilled.
| Metric Name | Description | Type | Tags | Unit | Status |
|:-|:-|:-|:-|:-|:-|
| request_concurrency | Concurrent requests that are routed to Activator
These are requests reported by the concurrency reporter which may not be done yet.
This is the average concurrency over a reporting period | Gauge | configuration_name
container_name
namespace_name
pod_name
revision_name
service_name | Dimensionless | Stable |
| request_count | The number of requests that are routed to Activator.
These are requests that have been fulfilled from the activator handler. | Counter | configuration_name
container_name
namespace_name
pod_name
response_code
response_code_class
revision_name
service_name | Dimensionless | Stable |
| request_latencies | The response time in millisecond for the fulfilled routed requests | Histogram | configuration_name
container_name
namespace_name
pod_name
response_code
response_code_class
revision_name
service_name | Milliseconds | Stable |
Autoscaler¶
Autoscaler component exposes a number of metrics related to its decisions per revision.
For example at any given time user can monitor the desired pods the Autoscaler wants to allocate for
a service, the average number of requests per second during the stable window, whether autoscaler is in panic mode (KPA) etc.
To read more about how autoscaler works check here.
| Metric Name | Description | Type | Tags | Unit | Status |
|:-|:-|:-|:-|:-|:-|
| desired_pods | Number of pods autoscaler wants to allocate | Gauge | configuration_name
namespace_name
revision_name
service_name | Dimensionless | Stable |
| excess_burst_capacity | Excess burst capacity overserved over the stable window | Gauge | configuration_name
namespace_name
revision_name
service_name | Dimensionless | Stable |
| stable_request_concurrency | Average of requests count per observed pod over the stable window | Gauge | configuration_name
namespace_name
revision_name
service_name | Dimensionless | Stable |
| panic_request_concurrency | Average of requests count per observed pod over the panic window | Gauge | configuration_name
namespace_name
revision_name
service_name | Dimensionless | Stable |
| target_concurrency_per_pod | The desired number of concurrent requests for each pod | Gauge | configuration_name
namespace_name
revision_name
service_name | Dimensionless | Stable |
| stable_requests_per_second | Average requests-per-second per observed pod over the stable window | Gauge | configuration_name
namespace_name
revision_name
service_name | Dimensionless | Stable |
| panic_requests_per_second | Average requests-per-second per observed pod over the panic window | Gauge | configuration_name
namespace_name
revision_name
service_name | Dimensionless | Stable |
| target_requests_per_second | The desired requests-per-second for each pod | Gauge | configuration_name
namespace_name
revision_name
service_name | Dimensionless | Stable |
| panic_mode | 1 if autoscaler is in panic mode, 0 otherwise | Gauge | configuration_name
namespace_name
revision_name
service_name | Dimensionless | Stable |
| requested_pods | Number of pods autoscaler requested from Kubernetes | Gauge | configuration_name
namespace_name
revision_name
service_name | Dimensionless | Stable |
| actual_pods | Number of pods that are allocated currently in ready state | Gauge | configuration_name
namespace_name
revision_name
service_name | Dimensionless | Stable |
| not_ready_pods | Number of pods that are not ready currently | Gauge | configuration_name=
namespace_name=
revision_name
service_name | Dimensionless | Stable |
| pending_pods | Number of pods that are pending currently | Gauge | configuration_name
namespace_name
revision_name
service_name | Dimensionless | Stable |
| terminating_pods | Number of pods that are terminating currently | Gauge | configuration_name
namespace_name
revision_name
service_name
| Dimensionless | Stable |
Controller¶
The following metrics are emitted by any component that implements a controller logic. The metrics show details about the reconciliation operations and the workqueue behavior on which reconciliation requests are enqueued.
| Metric Name | Description | Type | Tags | Unit | Status |
|---|---|---|---|---|---|
| work_queue_depth | Depth of the work queue | Gauge | reconciler | Dimensionless | Stable |
| reconcile_count | Number of reconcile operations | Counter | reconciler success |
Dimensionless | Stable |
| reconcile_latency | Latency of reconcile operations | Histogram | reconciler success |
Milliseconds | Stable |
| workqueue_adds_total | Total number of adds handled by workqueue | Counter | name | Dimensionless | Stable |
| workqueue_depth | Current depth of workqueue | Gauge | reconciler | Dimensionless | Stable |
| workqueue_queue_latency_seconds | How long in seconds an item stays in workqueue before being requested | Histogram | name | Seconds | Stable |
| workqueue_retries_total | Total number of retries handled by workqueue | Counter | name | Dimensionless | Stable |
| workqueue_work_duration_seconds | How long in seconds processing an item from a workqueue takes. | Histogram | name | Seconds | Stable |
| workqueue_unfinished_work_seconds | How long in seconds the outstanding workqueue items have been in flight (total). | Histogram | name | Seconds | Stable |
| workqueue_longest_running_processor_seconds | How long in seconds the longest outstanding workqueue item has been in flight | Histogram | name | Seconds | Stable |
Webhook¶
Webhook metrics report useful info about operations eg. CREATE on Serving resources and if admission was allowed.
For example if a big number of operations fail this could be an issue with the submitted user resource.
| Metric Name | Description | Type | Tags | Unit | Status |
|:-|:-|:-|:-|:-|:-|
| request_count | The number of requests that are routed to webhook | Counter | admission_allowed
kind_group
kind_kind
kind_version
request_operation
resource_group
resource_namespace
resource_resource
resource_version | Dimensionless | Stable |
| request_latencies | The response time in milliseconds | Histogram | admission_allowed
kind_group
kind_kind
kind_version
request_operation
resource_group
resource_namespace
resource_resource
resource_version | Milliseconds | Stable |
Go Runtime - memstats¶
Each Knative Serving control plane process emits a number of Go runtime memory statistics (shown next).
As a baseline for monitoring purproses, user could start with a subset of the metrics: current allocations (go_alloc), total allocations (go_total_alloc), system memory (go_sys), mallocs (go_mallocs), frees (go_frees) and garbage collection total pause time (total_gc_pause_ns), next gc target heap size (go_next_gc) and number of garbage collection cycles (num_gc).
| Metric Name | Description | Type | Tags | Unit | Status |
|:-|:-|:-|:-|:-|:-|
| go_alloc | The number of bytes of allocated heap objects (same as heap_alloc) | Gauge | name | Dimensionless | Stable |
| go_total_alloc | The cumulative bytes allocated for heap objects | Gauge | name | Dimensionless | Stable |
| go_sys | The total bytes of memory obtained from the OS | Gauge | name | Dimensionless | Stable |
| go_lookups | The number of pointer lookups performed by the runtime | Gauge | name | Dimensionless | Stable |
| go_mallocs | The cumulative count of heap objects allocated | Gauge | name | Dimensionless | Stable |
| go_frees | The cumulative count of heap objects freed | Gauge | name | Dimensionless | Stable |
| go_heap_alloc | The number of bytes of allocated heap objects | Gauge | name | Dimensionless | Stable |
| go_heap_sys | The number of bytes of heap memory obtained from the OS | Gauge | name | Dimensionless | Stable |
| go_heap_idle | The number of bytes in idle (unused) spans | Gauge | name | Dimensionless | Stable |
| go_heap_in_use | The number of bytes in in-use spans | Gauge | name | Dimensionless | Stable |
| go_heap_released | The number of bytes of physical memory returned to the OS | Gauge | name | Dimensionless | Stable |
| go_heap_objects | The number of allocated heap objects | Gauge | name | Dimensionless | Stable |
| go_stack_in_use | The number of bytes in stack spans | Gauge | name | Dimensionless | Stable |
| go_stack_sys | The number of bytes of stack memory obtained from the OS | Gauge | name | Dimensionless | Stable |
| go_mspan_in_use | The number of bytes of allocated mspan structures | Gauge | name | Dimensionless | Stable |
| go_mspan_sys | The number of bytes of memory obtained from the OS for mspan structures | Gauge | name | Dimensionless | Stable |
| go_mcache_in_use | The number of bytes of allocated mcache structures | Gauge | name | Dimensionless | Stable |
| go_mcache_sys | The number of bytes of memory obtained from the OS for mcache structures | Gauge | name | Dimensionless | Stable |
| go_bucket_hash_sys | The number of bytes of memory in profiling bucket hash tables. | Gauge | name | Dimensionless | Stable |
| go_gc_sys | The number of bytes of memory in garbage collection metadata | Gauge | name | Dimensionless | Stable |
| go_other_sys | The number of bytes of memory in miscellaneous off-heap runtime allocations | Gauge | name | Dimensionless | Stable |
| go_next_gc | The target heap size of the next GC cycle | Gauge | name | Dimensionless | Stable |
| go_last_gc | The time the last garbage collection finished, as nanoseconds since 1970 (the UNIX epoch) | Gauge | name | Nanoseconds | Stable |
| go_total_gc_pause_ns | The cumulative nanoseconds in GC stop-the-world pauses since the program started | Gauge | name | Nanoseconds | Stable |
| go_num_gc | The number of completed GC cycles. | Gauge | name | Dimensionless | Stable |
| go_num_forced_gc | The number of GC cycles that were forced by the application calling the GC function. | Gauge | name | Dimensionless | Stable |
| go_gc_cpu_fraction | The fraction of this program's available CPU time used by the GC since the program started | Gauge | name | Dimensionless | Stable |
NOTE: name tag is empty.
Developer - User Services¶
Every Knative service has a proxy container that proxies the connections to the application container. A number of metrics are reported for the queue peroxy performance. Using the following metrics application developers, devops and others, could measure if requests are queued at the proxy side (need for backpressure) and what is the actual delay in serving requests at the application side.
Queue proxy¶
Requests endpoint
| Metric Name | Description | Type | Tags | Unit | Status |
|---|---|---|---|---|---|
| revision_request_count | The number of requests that are routed to queue-proxy | Counter | configuration_name container_name namespace_name pod_name response_code response_code_class revision_name service_name |
Dimensionless | Stable |
| revision_request_latencies | The response time in millisecond | Histogram | configuration_name container_name namespace_name pod_name response_code response_code_class revision_name service_name |
Milliseconds | Stable |
| revision_app_request_count | The number of requests that are routed to user-container | Counter | configuration_name container_name namespace_name pod_name response_code response_code_class revision_name service_name |
Dimensionless | Stable |
| revision_app_request_latencies | The response time in millisecond | Histogram | configuration_name namespace_name pod_name response_code response_code_class revision_name service_name |
Milliseconds | Stable |
| revision_queue_depth | The current number of items in the serving and waiting queue, or not reported if unlimited concurrency | Gauge | configuration_name event-display container_name namespace_name pod_name response_code_class revision_name service_name |
Dimensionless | Stable |