Yandex Monitoring metric reference

Written by

Updated at November 27, 2025

CPU metrics
Disk metrics
Disk operation metrics
RAM metrics
Network metrics
Service metrics
Other metrics

This section describes Managed Service for Apache Kafka® metrics delivered to Monitoring.

The name label contains the metric name.

Labels shared by all Managed Service for Apache Kafka® metrics:

Label	Value
service	Service ID: `managed-kafka`
resource_type	Resource type: `cluster`
resource_id	Cluster ID
host	Host FQDN
node	Broker type: `leader`, `follower`, or `replica`
subcluster_name	Subcluster type: `zookeeper_subcluster` or `kafka_subcluster`

CPU metrics

CPU core workload.

Name Type, units	Description
`cpu.fraction` `DGAUGE`, %	Guaranteed vCPU performance
`cpu.guarantee` `DGAUGE`, count	Guaranteed number of cores
`cpu.limit` `DGAUGE`, count	Maximum number of cores in use
`cpu.guest` `DGAUGE`, %	CPU core usage, `guest` usage type
`cpu.idle` `DGAUGE`, %	CPU core usage, `idle` usage type
`cpu.iowait` `DGAUGE`, %	CPU core usage, `iowait` usage type
`cpu.irq` `DGAUGE`, %	CPU core usage, `irq` usage type
`cpu.nice` `DGAUGE`, %	CPU core usage, `nice` usage type
`cpu.softirq` `DGAUGE`, %	CPU core usage, `softirq` usage type
`cpu.steal` `DGAUGE`, %	CPU core usage, `steal` usage type
`cpu.system` `DGAUGE`, %	CPU core usage, `system` usage type
`cpu.user` `DGAUGE`, %	CPU core usage, `user` usage type
`load.avg_15min` `DGAUGE`, %	Average load over 15 minutes
`load.avg_1min` `DGAUGE`, %	Average load over one minute
`load.avg_5min` `DGAUGE`, %	Average load over five minutes

Disk metrics

Name Type, units	Description
`disk.free_bytes` `DGAUGE`, bytes	Free space
`disk.free_inodes` `DGAUGE`, count	Free inodes
`disk.total_bytes` `DGAUGE`, bytes	Available space
`disk.total_inodes` `DGAUGE`, count	Available inodes
`disk.used_bytes` `DGAUGE`, bytes	Used space
`disk.used_inodes` `DGAUGE`, count	Used inodes

Disk operation metrics

Name Type, units	Description
`io.avg_read_time` `DGAUGE`, milliseconds	Average disk read time
`io.avg_write_time` `DGAUGE`, milliseconds	Average disk write time
`io.disk*.avg_read_time` `DGAUGE`, milliseconds	Average read time for a given disk
`io.disk*.avg_write_time` `DGAUGE`, milliseconds	Average write time for a given disk
`io.disk*.read_bytes` `DGAUGE`, bytes per second	Read speed for a given disk
`io.disk*.read_count` `DGAUGE`, operations per second	Number of reads per second for a given disk
`io.disk*.read_merged_count` `DGAUGE`, operations per second	Number of merged read operations per second for a given disk
`io.disk*.utilization` `DGAUGE`, %	Utilization of a given disk; disabled for network drives.
`io.disk*.write_bytes` `DGAUGE`, bytes per second	Write speed for a given disk
`io.disk*.write_count` `DGAUGE`, operations per second	Number of writes per second for a given disk
`io.disk*.write_merged_count` `DGAUGE`, operations per second	Number of merged write operations per second for a given disk
`io.read_bytes` `DGAUGE`, bytes per second	Disk read rate
`io.read_count` `DGAUGE`, operations per second	Number of read operations per second
`io.read_merged_count` `DGAUGE`, operations per second	Number of merged read operations per second
`io.utilization` `DGAUGE`, %	Disk utilization disabled for network drives
`io.write_bytes` `DGAUGE`, bytes per second	Disk write speed
`io.write_count` `DGAUGE`, operations per second	Number of writes per second
`io.write_merged_count` `DGAUGE`, operations per second	Number of merged write operations per second

RAM metrics

Name Type, units	Description
`mem.guarantee_bytes` `DGAUGE`, bytes	Guaranteed memory allocation
`mem.limit_bytes` `DGAUGE`, bytes	Memory limit
`mem.active_bytes` `DGAUGE`, bytes	Active resident memory (frequently accessed and released when absolutely necessary)
`mem.available_bytes` `DGAUGE`, bytes	RAM usage, `available` usage type
`mem.buffers_bytes` `DGAUGE`, bytes	RAM usage, `buffers` usage type
`mem.cached_bytes` `DGAUGE`, bytes	RAM usage, `cached` usage type
`mem.free_bytes` `DGAUGE`, bytes	Amount of free RAM available, excluding `mem.buffers_bytes` and `mem.cached_bytes`
`mem.shared_bytes` `DGAUGE`, bytes	RAM usage, `shared` usage type
`mem.total_bytes` `DGAUGE`, bytes	RAM usage, `total` usage type
`mem.used_bytes` `DGAUGE`, bytes	Amount of RAM currently used by running processes

Network metrics

Name Type, units	Description
`net.bytes_recv` `DGAUGE`, bytes per second	Network data receive rate
`net.bytes_sent` `DGAUGE`, bytes per second	Network data transmit rate
`net.dropin` `DGAUGE`, count	Dropped receive packets
`net.dropout` `DGAUGE`, count	Dropped transmit packets
`net.errin` `DGAUGE`, count	Receive error count
`net.errout` `DGAUGE`, count	Transmit error count
`net.packets_recv` `DGAUGE`, packets per second	Network packet receive rate
`net.packets_sent` `DGAUGE`, packets per second	Network packet transmit rate

Service metrics

Name Type, units	Description
`kafka_controller_ControllerStats_LeaderElectionRateAndTimeMs` `DGAUGE`, milliseconds	Leader broker switch rate per unit of time. In a normal state, it is `0`. Its value may increase during maintenance, which does not indicate a problem. Additional labels: `quantile`.
`kafka_controller_KafkaController_ActiveControllerCount` `DGAUGE`, count	Number of active controllers.
`kafka_controller_KafkaController_GlobalTopicCount` `DGAUGE`, count	Number of topics
`kafka_controller_KafkaController_OfflinePartitionsCount` `DGAUGE`, count	Number of offline partitions.
`kafka_controller_KafkaController_PreferredReplicaImbalanceCount` `DGAUGE`, count	Imbalance count in the preferred replica. In a normal state, it is `0`.
`kafka_group_topic_partition_lag` `DGAUGE`, count	Message lag: Difference between the consumer offset and the partition's latest offset.
`kafka_group_topic_partition_offset` `DGAUGE`, count	Partition's current consumer group offset.
`kafka_log_Log_LogStartOffset` `DGAUGE`, count	Partition's first offset.
`kafka_log_Log_LogEndOffset` `DGAUGE`, count	Partition's last offset.
`kafka_log_Log_Size` `DGAUGE`, bytes	Disk’s partition size.
`kafka_host_count` `DGAUGE`, count	Number of hosts in the cluster
`kafka_is_alive` `DGAUGE`, `0`/`1`	Broker health indicator. The metric calculation algorithm depends on whether there are any highly available topics (further referred to as HA topics) and which state their partition leaders have: If the broker has at least one HA topic, the algorithm checks only HA topics. The number of replicas each HA topic has must be equal to or greater than the `Minimum number of in-sync replicas`. If the broker has no HA topics, the algorithm checks all topics. The number of replicas each topic has must be equal to or greater than the `Minimum number of in-sync replicas`. For more information about the `Minimum number of in-sync replicas`, see the Apache Kafka® settings section. It can be either `1` if a broker is alive or `0` if it is not.
`kafka_network_RequestChannel_RequestQueueSize` `DGAUGE`, count	Number of enqueued requests
`kafka_network_RequestMetrics_Errors` `DGAUGE`, count	Number of errors. Additional labels: `request`.
`kafka_network_RequestMetrics_LocalTimeMs` `DGAUGE`, milliseconds	Time it takes the leader broker to process a request. Additional labels: `request` and `quantile`.
`kafka_network_RequestMetrics_MessageConversionsTimeMs` `DGAUGE`, milliseconds	Message format conversion time. Additional labels: `request` and `quantile`.
`kafka_network_RequestMetrics_RemoteTimeMs` `DGAUGE`, milliseconds	Follower broker wait time. Additional labels: `request` and `quantile`.
`kafka_network_RequestMetrics_RequestQueueTimeMs` `DGAUGE`, milliseconds	Request queue wait time. Additional labels: `request` and `quantile`.
`kafka_network_RequestMetrics_Requests` `DGAUGE`, count	Number of requests. Additional labels: `request`.
`kafka_network_RequestMetrics_ResponseQueueTimeMs` `DGAUGE`, milliseconds	Response queue wait time. Additional labels: `request` and `quantile`.
`kafka_network_RequestMetrics_ResponseSendTimeMs` `DGAUGE`, milliseconds	Response send time. Additional labels: `request` and `quantile`.
`kafka_network_RequestMetrics_TotalTimeMs` `DGAUGE`, milliseconds	Total request execution time. Additional labels: `request` and `quantile`.
`kafka_network_SocketServer_NetworkProcessorAvgIdlePercent` `DGAUGE`, %	Average network processor idle percentage. Its value ranges from `0` (fully utilized) to `1` (completely idle).
`kafka_server_BrokerTopicMetrics_BytesIn` `DGAUGE`, bytes	Incoming data size
`kafka_server_BrokerTopicMetrics_BytesOut` `DGAUGE`, bytes	Outgoing data size
`kafka_server_BrokerTopicMetrics_FailedFetchRequests` `DGAUGE`, count	Number of requests received with errors
`kafka_server_BrokerTopicMetrics_FailedProduceRequests` `DGAUGE`, count	Number of requests processed with errors
`kafka_server_BrokerTopicMetrics_MessagesIn` `DGAUGE`, count	Number of written messages
`kafka_server_BrokerTopicMetrics_ReplicationBytesIn` `DGAUGE`, bytes	Replicated data size
`kafka_server_KafkaRequestHandlerPool_RequestHandlerAvgIdlePercent_count` `DGAUGE`, %	Average request handler idle percentage. Its value ranges from `0` (fully utilized) to `1` (completely idle).
`kafka_server_KafkaServer_BrokerState` `DGAUGE`	Broker state: 0: `Not Running` 1: `Starting` 2: `Recovering from Unclean Shutdown` 3: `Running as Broker` 4: `Running as Controller` 5: `Pending Controlled ShutdownStates` 6: `Broker Shutting Down`
`kafka_server_ReplicaFetcherManager_MaxLag` `DGAUGE`, count	Maximum lag of message replication between the follower and leader brokers. Additional labels: `clientId`.
`kafka_server_ReplicaManager_LeaderCount` `DGAUGE`, count	Number of partitions led by the broker
`kafka_server_ReplicaManager_OfflineReplicaCount` `DGAUGE`, count	Number of partitions with no leader broker. These partitions do not support message writes or reads.
`kafka_server_ReplicaManager_PartitionCount` `DGAUGE`, count	Number of partitions per broker
`kafka_server_ReplicaManager_ReassigningPartitions` `DGAUGE`, count	Number of partitions with the leader being reassigned
`kafka_server_ReplicaManager_UnderMinIsrPartitionCount` `DGAUGE`, count	Number of partitions with in-sync replica (ISR) count below the set minimum
`kafka_server_ReplicaManager_UnderReplicatedPartitions` `DGAUGE`, count	Number of partitions with ISR count below the replication factor
`kafka_server_ZooKeeperClientMetrics_ZooKeeperRequestLatencyMs` `DGAUGE`, milliseconds	Request latency in ZooKeeper. Additional labels: `quantile`.
`kafka_shard_count` `DGAUGE`, count	Number of active shards
`kafka_topic_partition_high_water_mark` `DGAUGE`, count	Partition max offset
`kafka_topic_partition_low_water_mark` `DGAUGE`, count	Partition min offset

Note

This section lists only basic Managed Service for Apache Kafka® metrics delivered to Monitoring. For information on all Managed Service for Apache Kafka® metrics, see the official documentation.

Other metrics

Name
Type, units

Description

can_read
DGAUGE, 0/1

Host read access indicator. The metric calculation algorithm depends on whether there are any highly available topics (further referred to as HA topics) and which state their partition leaders have:

If the host has at least one HA topic, the algorithm checks only HA topics. Each HA topic’s partition leader must be online.
If the host has no HA topics, the algorithm checks all topics. Each topic’s partition leader must be online.

It can be either 1 if the host is available for reads or 0 if it is not.

can_write
DGAUGE, 0/1

Host write access indicator. The metric calculation algorithm depends on whether there are any highly available topics (further referred to as HA topics) and which state their partition leaders have:

If the host has at least one HA topic, the algorithm checks only HA topics. The number of replicas each HA topic has must be equal to or greater than the Minimum number of in-sync replicas.
If the host has no HA topics, the algorithm checks all topics. The number of replicas each topic has must be equal to or greater than the Minimum number of in-sync replicas.

For more information about the Minimum number of in-sync replicas, see the Apache Kafka® settings section.

Additionally, the storage is checked for available space. It should be more than 5%. If there is not enough space, the host is unavailable for writes.

It can be either 1 if the host is available for writes or 0 if it is not.

Yandex Monitoring metric reference

CPU metricsCPU metrics

Disk metricsDisk metrics

Disk operation metricsDisk operation metrics

RAM metricsRAM metrics

Network metricsNetwork metrics

Service metricsService metrics

Other metricsOther metrics