Public Metrics

This section provides reference descriptions about the public metrics exported from Redpanda’s /public_metrics endpoint.

In a live system, Redpanda metrics are exported only for features that are in use. For example, Redpanda does not export metrics for consumer groups if no groups are registered.

To see the available public metrics in your system, query the /public_metrics endpoint:

curl http://<node-addr>:9644/public_metrics | grep "[HELP|TYPE]"

Use /public_metrics for your primary dashboards for system health.

Use /metrics for detailed analysis and debugging.

Cluster metrics

redpanda_cluster_brokers

Number of configured, fully commissioned brokers in a cluster.

Type: gauge

Usage: Create an alert for when this gauge dips below a steady-state threshold, as a node(s) has become unresponsive.


redpanda_cluster_controller_log_limit_requests_available_rps

Limit on the available requests per second (RPS) for a cluster controller log.

Type: gauge

Labels:

  • redpanda_cmd_group=("move_operations" | "topic_operations" | "configuration_operations" | "node_management_operations" | "acls_and_users_operations")


redpanda_cluster_controller_log_limit_requests_dropped

Number of requests dropped by a cluster controller log due to exceeding redpanda_cluster_controller_log_limit_requests_available_rps.

Type: counter

Labels:

  • redpanda_cmd_group=("move_operations" | "topic_operations" | "configuration_operations" | "node_management_operations" | "acls_and_users_operations")

Usage: When this counter increases, it indicates that the controller is dropping requests.


redpanda_cluster_features_enterprise_license_expiry_sec

Number of seconds remaining until the Enterprise Edition license expires.

Type: gauge

Usage: Use this metric to set up alerts or dashboards to track the license expiration status and ensure timely renewal.


redpanda_cluster_non_homogenous_fips_mode

Number of brokers that have a non-homogenous FIPS mode value.

Type: gauge


redpanda_cluster_partition_moving_from_node

Number of partition replicas in the cluster that are currently being removed from a node.

Type: gauge

Usage: When this gauge is non-zero, determine whether there is an expected or unexpected reassignment of partitions that is causing movement of partition replicas.


redpanda_cluster_partition_moving_to_node

Number of partition replicas in the cluster that are currently being added or moved to a node.

Type: gauge

Usage: When this gauge is non-zero, determine whether there is an expected or unexpected reassignment of partitions that is causing movement of partition replicas.


redpanda_cluster_partition_node_cancelling_movements

During a partition movement cancellation operation, the number of partition replicas that were being moved that now need to be cancelled.

Type: gauge

Usage: When this gauge is non-zero, determine whether there is an expected or unexpected reassignment of partitions that is causing movement of partition replicas.


redpanda_cluster_partitions

Number of partitions managed by a cluster. Includes partitions of the controller topic, but not replicas.

Type: gauge


redpanda_cluster_topics

Number of topics in a cluster.

Type: gauge


redpanda_cluster_unavailable_partitions

Number of unavailable partitions (the partitions that lack quorum among their replica group) in the cluster.

Type: gauge

Usage: When this gauge is non-zero, it indicates that a partition(s) doesn’t have quorum and thus doesn’t have an active leader. To mitigate, consider increasing the number of brokers and the partition replication factor.


redpanda_cluster_latest_cluster_metadata_manifest_age

The amount of time in seconds since the last time Redpanda uploaded metadata files to Tiered Storage for your cluster. A value of 0 indicates metadata has not yet been uploaded.

When performing a whole cluster restore operation, the metadata for the new cluster will not have any changes made to a source cluster that is newer than this age.

Type: gauge

Usage: On a healthy system, this should not exceed the value set for cloud_storage_cluster_metadata_upload_interval_ms. You may consider setting an alert if this remains 0 for longer than 1.5 * cloud_storage_cluster_metadata_upload_interval_ms as that may indicate a configuration issue.

Related topics:

Infrastructure metrics

redpanda_cpu_busy_seconds_total

Total CPU busy time in seconds.

Type: counter


redpanda_io_queue_total_read_ops

Total read operations passed in the I/O queue.

Type: counter

Labels:

  • class=("default" | "compaction" | "raft")

  • ioshard

  • mountpoint

  • shard


redpanda_io_queue_total_write_ops

Total write operations passed in the I/O queue.

Type: counter

Labels:

  • class=("default" | "compaction" | "raft")

  • ioshard

  • mountpoint

  • shard


redpanda_memory_allocated_memory

Total allocated memory in bytes.

Type: gauge

Labels:

  • shard


redpanda_memory_available_memory

Total memory potentially available (free plus reclaimable memory) to a CPU shard (core), in bytes.

Type: gauge

Labels:

  • shard


redpanda_memory_available_memory_low_water_mark

The low watermark for available memory at process start.

Type: gauge

Labels:

  • shard


redpanda_memory_free_memory

Available memory in bytes.

Type: gauge

Labels:

  • shard


redpanda_rpc_active_connections

The total number of currently-active clients the RPC server on a given shard has connections to.

Type: gauge

Labels:

  • redpanda_server=("kafka" | "internal")


redpanda_rpc_request_errors_total

Number of RPC errors.

Type: counter

Labels:

  • redpanda_server=("kafka" | "internal")

Usage: When this counter increases, analyze the logged errors.


redpanda_rpc_request_latency_seconds

RPC latency in seconds.

Type: histogram

Labels:

  • redpanda_server=("kafka" | "internal")


redpanda_scheduler_runtime_seconds_total

Accumulated runtime of the task queue associated with a scheduling group.

Type: counter

Labels:

  • redpanda_scheduling_group=("admin" | "archival_upload" | "cache_background_reclaim" | "cluster" | "coproc" | "kafka" | "log_compaction" | "main" | "node_status" | "raft" | "raft_learner_recovery")

  • shard


redpanda_storage_disk_free_bytes

Available disk storage in bytes.

Type: gauge


redpanda_storage_disk_free_space_alert

Alert for low disk storage: 0-OK, 1-low space, 2-degraded.

Type: gauge


redpanda_storage_disk_total_bytes

Total size in bytes of attached storage.

Type: gauge


redpanda_uptime_seconds_total

Total CPU runtime (uptime) in seconds.

Type: gauge


Raft metrics

redpanda_node_status_rpcs_received

Number of node status RPCs received by a node.

Type: gauge


redpanda_node_status_rpcs_sent

Number of node status RPCs sent by a node.

Type: gauge


redpanda_node_status_rpcs_timed_out

Number of timed out node status RPCs from a node.

Type: gauge


Service metrics

redpanda_pandaproxy_request_latency_seconds

Latency in seconds of the request indicated by the label in HTTP Proxy. The measurement includes the time spent waiting for resources to become available, processing the request, and dispatching the response.

Type: histogram


redpanda_schema_registry_request_errors_total

Total number of Schema Registry errors.

Type: counter

Labels:

  • redpanda_status=("5xx" | "4xx" | "3xx")


redpanda_schema_registry_request_latency_seconds

Latency of the request indicated by the label in the Schema Registry. The measurement includes the time spent waiting for resources to become available, processing the request, and dispatching the response.

Type: histogram


Partition metrics

redpanda_kafka_max_offset

The high watermark offset of a partition. Suitable for calculating consumer group lag.

Type: gauge

Labels:

  • redpanda_namespace

  • redpanda_partition

  • redpanda_topic

Related topics:


redpanda_kafka_under_replicated_replicas

Number of replicas in the partition that are live but not at the latest offset, redpanda_kafka_max_offset.

Type: gauge

Labels:

  • redpanda_namespace

  • redpanda_partition

  • redpanda_topic


redpanda_raft_recovery_partition_movement_available_bandwidth

Bandwidth available for partition movement, in bytes per sec.

Type: gauge

Labels:

  • shard


Topic metrics


redpanda_cluster_partition_schema_id_validation_records_failed

Number of records that failed schema ID validation.

Type: counter


redpanda_kafka_partitions

Configured number of partitions for a topic.

Type: gauge

Labels: -redpanda_namespace -redpanda_topic


redpanda_kafka_records_fetched_total

Total number of records fetched.

Type: counter


redpanda_kafka_records_produced_total

Total number of records produced.

Type: counter


redpanda_kafka_replicas

The number of configured replicas per topic.

Type: gauge

Labels:

  • redpanda_namespace

  • redpanda_topic


redpanda_kafka_request_bytes_total

Total number of bytes produced or consumed per topic.

Type: counter

Labels:

  • redpanda_namespace

  • redpanda_topic

  • redpanda_request=("produce" | "consume")


redpanda_raft_leadership_changes

Number of leadership changes across all partitions of a given topic.

Type: counter

Labels:

  • redpanda_namespace

  • redpanda_topic


Broker metrics

redpanda_kafka_request_latency_seconds

Latency of produce/consume requests per broker. This duration measures from when a request is initiated on the partition to when the response is fulfilled.

Type: histogram

Labels:

  • redpanda_request=("produce" | "consume")


redpanda_kafka_quotas_client_quota_throttle_time

Client quota throttling delay, in seconds, per rule and quota type.

Type: histogram

Labels:

  • quota_rule=("not_applicable" | "kafka_client_default" | "cluster_client_default" | "kafka_client_prefix" | "cluster_client_prefix" | "kafka_client_id")`

  • quota_type=("produce_quota" | "fetch_quota" | "partition_mutation_quota")


redpanda_kafka_quotas_client_quota_throughput

Client quota throughput per rule and quota type.

Type: histogram

Labels:

  • quota_rule=("not_applicable" | "kafka_client_default" | "cluster_client_default" | "kafka_client_prefix" | "cluster_client_prefix" | "kafka_client_id")

  • quota_type=("produce_quota" | "fetch_quota" | "partition_mutation_quota")


Consumer group metrics

redpanda_kafka_consumer_group_committed_offset

Committed offset of a consumer group.

Type: gauge

Labels:

  • group

  • topic

  • partition


redpanda_kafka_consumer_group_consumers

Number of consumers in a consumer group.

Type: gauge

Labels:

  • group


redpanda_kafka_consumer_group_topics

Number of topics in a consumer group.

Type: gauge

Labels:

  • group


REST proxy metrics

redpanda_rest_proxy_request_errors_total

Total number of REST proxy server errors.

Type: counter

Labels:

  • redpanda_status("5xx" | "4xx" | "3xx")


redpanda_rest_proxy_request_latency_seconds_bucket

Internal latency of REST proxy requests.

Type: histogram


Application metrics

redpanda_application_build

Redpanda build information.

Type: gauge

Labels:

  • redpanda_revision=<redpanda-revision-ID>

  • redpanda_version=<redpanda-version-number>


redpanda_application_fips_mode

Identifies whether or not Redpanda is running in FIPS mode. Possible values:

  • 0 = disabled

  • 1 = permissive

  • 2 = enabled

Type: gauge


redpanda_application_uptime_seconds_total

Redpanda application uptime in seconds.

Type gauge


Cloud metrics

redpanda_cloud_client_backoff

Total number of requests that backed off.

Type: counter

Labels:

  • S3

    • redpanda_endpoint

    • redpanda_region

  • Azure Blob Storage (ABS)

    • redpanda_endpoint

    • redpanda_storage_account


redpanda_cloud_client_download_backoff

Total number of download requests that backed off.

Type: counter

Labels:

  • S3

    • redpanda_endpoint

    • redpanda_region

  • Azure Blob Storage (ABS)

    • redpanda_endpoint

    • redpanda_storage_account


redpanda_cloud_client_downloads

Total number of requests that downloaded an object from cloud storage.

Type: counter

Labels:

  • S3

    • redpanda_endpoint

    • redpanda_region

  • Azure Blob Storage (ABS)

    • redpanda_endpoint

    • redpanda_storage_account


redpanda_cloud_client_not_found

Total number of requests for which the object was not found.

Type: counter

Labels:

  • S3

    • redpanda_endpoint

    • redpanda_region

  • Azure Blob Storage (ABS)

    • redpanda_endpoint

    • redpanda_storage_account


redpanda_cloud_client_upload_backoff

Total number of upload requests that backed off.

Type: counter

Labels:

  • S3

    • redpanda_endpoint

    • redpanda_region

  • Azure Blob Storage (ABS)

    • redpanda_endpoint

    • redpanda_storage_account


redpanda_cloud_client_uploads

Total number of requests that uploaded an object to cloud storage.

Type: counter

Labels:

  • S3

    • redpanda_endpoint

    • redpanda_region

  • Azure Blob Storage (ABS)

    • redpanda_endpoint

    • redpanda_storage_account


repdanda_cloud_storage_cloud_log_size

The amount of data, in bytes, that exists in Tiered Storage and is accessible by Kafka. Increases every time Redpanda offloads a segment to Tiered Storage, and decreases every time compaction or retention deletes data.

The redpanda_namespace label supports the following options for this metric:

  • kafka - User topics

  • kafka_internal - Internal Kafka topic, such as consumer groups

  • redpanda - Redpanda-only internal data

Type: gauge

Usage: This metric reports the log size for each topic and partition in your cluster

Labels:

  • redpanda_namespace=("kafka" | "kafka_internal" | "redpanda")

  • redpanda_topic

  • redpanda_partition

TLS metrics

redpanda_tls_truststore_expires_at_timestamp_seconds

The expiration time, represented as a Unix-style timestamp, of the shortest-lived certificate authority (CA) in the installed CA chain. This lists all resources that have at least one CA configured.

Type: gauge

Labels:

  • area

  • detail

Usage: This metric is recorded for all resources that may have a TLS certificate installed. The area and detail labels provide a hierarchical way of identifying a single resource you need information on.


redpanda_tls_certificate_expires_at_timestamp_seconds

The expiration time, represented as a Unix-style timestamp, for the shortest-lived installed certificate. While more than one certificate may be installed for a resource, this provides only the timestamp of the next certificate to expire. This lists all resources that have at least one certificate installed.

Type: gauge

Labels:

  • area

  • detail

Usage: This metric is recorded for all resources that may have a TLS certificate installed. The area and detail labels provide a hierarchical way of identifying a single resource you need information on.


redpanda_tls_certificate_serial

The serial number of the shortest-lived installed certificate. These serial numbers are technically unbounded. As such, this will return the least significant 4B digits of the integer representation certificate’s serial number. While more than one certificate may be installed for a resource, this provides only the serial number of the next certificate to expire. This lists all resources that have at least one certificate installed.

Type: gauge

Labels:

  • area

  • detail

Usage: This metric is recorded for all resources that may have a TLS certificate installed. The area and detail labels provide a hierarchical way of identifying a single resource you need information on.


redpanda_tls_loaded_at_timestamp_seconds

The last time, represented as a Unix-style timestamp, that at least one TLS certificate was loaded for a resource. This lists all resources that have at least one certificate installed.

Type: gauge

Labels:

  • area

  • detail

Usage: This metric is recorded for all resources that may have a TLS certificate installed. The area and detail labels provide a hierarchical way of identifying a single resource you need information on.


redpanda_tls_certificate_valid

Indicates whether a given resource has at least one valid installed TLS certificate. Querying this will provide a check on all resources with at least one certificate installed. The value of this gauge is either 1 to indicate presence of a valid certificate or 0 to indicate the absence of a valid certificate.

Type: gauge

Labels:

  • area

  • detail

Usage: This metric is recorded for all resources that may have a TLS certificate installed. The area and detail labels provide a hierarchical way of identifying a single resource you need information on.


Data transforms metrics

redpanda_transform_execution_latency_sec

A histogram of the latency of transforming a single record, in seconds.

Type: histogram

Labels:

  • function_name


redpanda_transform_execution_errors

Data transform invocation errors.

Type: counter

Labels:

  • function_name


redpanda_wasm_engine_cpu_seconds_total

Total CPU time spent inside a WebAssembly function, in seconds.

Type: counter

Labels:

  • function_name


redpanda_wasm_engine_memory_usage

Amount of memory usage for a WebAssembly function.

Type: gauge

Labels:

  • function_name


redpanda_wasm_engine_max_memory

Maximum amount of memory for a WebAssembly function.

Type: gauge

Labels:

  • function_name


redpanda_wasm_binary_executable_memory_usage

The amount of executable memory usage for WebAssembly binaries.

Type: gauge


redpanda_transform_read_bytes

A counter for all the bytes that has been input to a transform.

Type: counter

Labels:

  • function_name


redpanda_transform_write_bytes

A counter for all the bytes that has been output from a transform.

Type: counter

Labels:

  • function_name


redpanda_transform_processor_lag

The amount of pending records on the input topic that have yet to be processed by the transform.

Type: gauge

Labels:

  • function_name


redpanda_transform_failures

A counter for each time that a processor encounters a failure.

Type: counter

Labels:

  • function_name


redpanda_transform_state

The number of transform processors in a specific state (running, inactive, errored).

Type: gauge

Labels:

  • function_name

  • state=("running" | "inactive" | "errored")


Cloud storage metrics

In a live system, Redpanda metrics are exported only for features that are in use. For example, Redpanda does not export metrics for consumer groups if no groups are registered.

To see the available public metrics in your system, query the /public_metrics endpoint:

curl http://<node-addr>:9644/public_metrics | grep "[HELP|TYPE]"
Cloud storage metrics are only available if you have:

redpanda_cloud_storage_cache_space_size_bytes

Sum of size of cached objects.

redpanda_cloud_storage_housekeeping_drains

Number of times upload housekeeping queue was drained.

redpanda_cloud_storage_spillover_manifests_materialized_bytes

Bytes of memory used for spilled manifests currently cached in memory.

redpanda_cloud_storage_cache_op_put

Number of objects written into cache.

redpanda_cloud_storage_segments

Total number of accounted topic segments in the cloud.

redpanda_cloud_storage_jobs_local_segment_reuploads

Number of segment reuploads from local data directory.

redpanda_cloud_storage_cache_trim_failed_trims

Number of times Redpanda could not free the expected amount of space, indicating possible bug or configuration issue.

redpanda_cloud_storage_cache_trim_exhaustive_trims

Number of times a fast trim could not free enough space and had to fall back to a slower exhaustive trim.

redpanda_cloud_storage_deleted_segments

Count of deleted remote segments.

redpanda_cloud_storage_segment_uploads_total

Successful data segment uploads.

redpanda_cloud_storage_active_segments

Number of remote log segments currently hydrated for read.

redpanda_cloud_storage_cache_trim_fast_trims

Number of times Redpanda trimmed the cache using the normal (fast) mode.

redpanda_cloud_storage_housekeeping_jobs_failed

Number of failed housekeeping jobs.

redpanda_cloud_storage_partition_readers_delayed

How many partition readers were delayed due to hitting reader limit. This indicates cluster is saturated with Tiered Storage reads.

redpanda_cloud_storage_segments_pending_deletion

Total number of topic segments pending deletion from the cloud.

redpanda_cloud_storage_housekeeping_rounds

Number of upload housekeeping rounds.

redpanda_cloud_storage_segment_readers_delayed

Number of segment readers delayed due to hitting reader limit. This indicates cluster is saturated with Tiered Storage reads.

redpanda_cloud_storage_cache_space_hwm_size_bytes

High watermark of sum of size of cached objects.

redpanda_cloud_storage_cache_space_hwm_files

High watermark of number of objects in cache.

redpanda_cloud_storage_cache_op_in_progress_files

Number of files that are being added to cache.

redpanda_cloud_storage_jobs_cloud_segment_reuploads

Number of segment reuploads from cloud storage sources (cloud storage cache or direct download from cloud storage).

redpanda_cloud_storage_jobs_manifest_reuploads

Number of manifest reuploads performed by all housekeeping jobs.

redpanda_cloud_storage_housekeeping_pauses

Number of times upload housekeeping was paused.

redpanda_cloud_storage_segment_index_uploads_total

Successful segment index uploads.

redpanda_cloud_storage_cache_op_miss

Number of failed get requests because of missing object in the cache.

redpanda_cloud_storage_errors_total

Number of transmit errors.

redpanda_cloud_storage_spillover_manifest_uploads_total

Successful spillover manifest uploads.

redpanda_cloud_storage_housekeeping_requests_throttled_average_rate

Average rate per shard of requests from the read and write path that were throttled by Tiered Storage.

redpanda_cloud_storage_jobs_segment_deletions

Number of segments deleted by all housekeeping jobs.

redpanda_cloud_storage_segment_materializations_delayed

Number of segment materializations delayed due to hitting reader limit. This indicates cluster is saturated with Tiered Storage reads.

redpanda_cloud_storage_jobs_metadata_syncs

Number of archival configuration updates performed by all housekeeping jobs.

redpanda_cloud_storage_housekeeping_jobs_completed

Number of successful housekeeping jobs.

redpanda_cloud_storage_readers

Number of segment read cursors for hydrated remote log segments.

redpanda_cloud_storage_partition_manifest_uploads_total

Successful partition manifest uploads.

redpanda_cloud_storage_limits_downloads_throttled_sum

Total amount of throttling applied to cloud storage downloads.

redpanda_cloud_storage_housekeeping_resumes

Number of times upload housekeeping was resumed.

redpanda_cloud_storage_cache_op_hit

Number of get requests for objects that are already in cache.

redpanda_cloud_storage_spillover_manifests_materialized_count

Number of spilled manifests currently cached in memory.

redpanda_cloud_storage_uploaded_bytes

Total number of uploaded bytes for the topic.

redpanda_cloud_storage_cache_space_files

Number of objects in cache.

redpanda_cloud_storage_housekeeping_jobs_skipped

Number of skipped housekeeping jobs.

redpanda_cloud_storage_partition_readers

Number of partition reader instances, based on the number of current fetch/timequery requests reading from Tiered Storage.