IBM Cloud Docs
Monitoring Event Streams service metrics by using IBM Cloud Monitoring

Monitoring Event Streams service metrics by using IBM Cloud Monitoring

IBM Cloud® Monitoring is a third-party cloud-native, and container-intelligence management system that you can include as part of your IBM Cloud architecture. Use it to gain operational visibility into the performance and health of your applications, services, and platforms. It offers administrators, DevOps teams, and developers full stack telemetry with advanced features to monitor and troubleshoot, define alerts, and design custom dashboards.

While you monitor service metrics with IBM Cloud Monitoring, Kafka clients (producers and consumers) have their own set of metrics ato monitor their performance and health.

Opting in to and enabling Event Streams service metrics

Event Streams service metrics can broadly be categorized into two different groups: Default and Enhanced.

Enabling default Event Streams service metrics

Before you can start to use Event Streams IBM Cloud Monitoring metrics, you must first opt in, and then enable platform metrics by completing the following steps:

  1. Enable platform metrics for Event Streams. For more information, see Enabling platform metrics.

    The owner of the account has full access to this metrics data. For more information about managing access for other users, see Getting started with IBM Cloud Monitoring - manage user access.

  2. To navigate from the Event Streams instance page to the Monitoring dashboard, click Actions on the instance page and select Monitoring.

    On your first usage, you might see a welcome wizard. To advance to the dashboard selection menu, select Next and then Skip on the Choosing an installation method page. Accept the prompts that follow. You can then select the IBM Event Streams or IBM Event Streams (Enterprise) dashboard, depending on the plan that you use.

Enabling enhanced Event Streams metrics

The enhanced Event Streams metrics consist of three groups; topic, partition and consumers. You can opt in to either one, two, or all. The metrics available are described in the topic, partition and consumers tables.

Enabling enhanced metrics introduces more global gauge metrics and therefore increases the costs.

Before you can start to use enhanced Event Streams metrics, you must first enable them by completing the following step:

  • Run the following command to update the service instance to start using enhanced metrics:

    ibmcloud resource service-instance-update <instance-name> -p '{"metrics":["topic","partition","consumers"]}'
    

When enhanced metrics are enabled, depending on the selection, the following new dashboards are available; IBM Event Streams(Topic), IBM Event Streams(Partitions) and IBM Event Streams(Consumers).

To opt out of enhanced metrics, run the following command:

ibmcloud resource service-instance-update <instance-name> -p '{"metrics":[]}'

Dashboards are available only after metrics started to be recorded; it might take a few minutes to initialize.

Event Streams service metrics cost information

Before you opt in to using Monitoring metrics, be aware of the cost of doing so. The estimated cost depends on the following considerations:

  • The Event Streams plan that you use.
  • How many unique time series are sent for each plan.
  • The number of topics that you created.
  • The number of partitions that you created.
  • Whether you have topics, partitions, consumers, or all enabled.

Enabling mirroring for Enterprise clusters introduces one more global gauge metric and an extra gauge metric per topic in the target cluster (with the target cluster already emitting metrics in accordance with the preceding table), therefore increasing the costs.

For more information, see Monitoring pricing.

Event Streams service metrics details

The following tables describe the specific metrics that are provided by Event Streams for each plan.

Service metrics available by service plan

Table 1. Metrics Available by Plan Names
Metric name Enterprise Lite Standard
Authentication failures Checkmark icon
Connected clients software name and version Checkmark icon
Consume message conversion time Checkmark icon
Estimated connected clients percentage Checkmark icon Checkmark icon Checkmark icon
Inactive consumer groups Checkmark icon
Instance bytes in per second Checkmark icon Checkmark icon Checkmark icon
Instance bytes out per second Checkmark icon Checkmark icon Checkmark icon
Missing SNI connections Checkmark icon
Number of offline partitions Checkmark icon
Number of partitions Checkmark icon Checkmark icon Checkmark icon
Number of topics Checkmark icon Checkmark icon Checkmark icon
Number of under in-sync replica partitions Checkmark icon
Produce message conversion time Checkmark icon
Rebalancing consumer groups Checkmark icon
Reserved disk space percentage Checkmark icon
Rest-producer requests per second Checkmark icon
Schema greatest version percentage Checkmark icon
Schema used percentage Checkmark icon
Stable consumer groups Checkmark icon
Topic bytes in per second Checkmark icon Checkmark icon Checkmark icon
Topic bytes out per second Checkmark icon Checkmark icon Checkmark icon
IAM ID bytes in per second Checkmark icon
IAM ID bytes out per second Checkmark icon
Used disk space percentage Checkmark icon
Instance utilization Checkmark icon

Service metrics available with mirroring enabled

Table 2. Metrics available for mirroring
Metric name Enterprise Lite Standard
Mirroring throughput Checkmark icon
Mirroring latency Checkmark icon

Enhanced service metrics available with topic enabled

Table 3. Metrics available for topic
Metric name Enterprise Lite Standard
Maximum partition retention percentage Checkmark icon
Topic size Checkmark icon

Enhanced service metrics available with consumers enabled

Table 4. Metrics available for consumers
Metric name Enterprise Lite Standard
Consumer groups lag Checkmark icon

Enhanced service metrics available with partitions enabled

Table 5. Metrics available for partitions
Metric name Enterprise Lite Standard
Message rate per partition Checkmark icon

This information is useful for detecting if the distribution of message activity across the partitions in a topic is unbalanced and if the number of partitions a topic is scaled appropriately.

Service metrics available with quotas enabled

Table 6. Metrics available for quotas
Metric name Enterprise Lite Standard
IAM ID bytes in quota used percentage Checkmark icon
IAM ID bytes out quota used percentage Checkmark icon

Kafka quotas use sampling to determine how long clients should be paused before they can send or receive more data. For unpredictable workloads, or configurations that result in quota decisions being made using only a few samples, you might observe the percentage quota used metric going above 100%.

Authentication failures

Incrementing count of the number of authentication failures

Table 7. Authentication failures metric metadata
Metadata Description
Metric Name ibm_eventstreams_kafka_authentication_failure_total
Metric Type counter
Value Type none
Segment By Service instance, Service instance name

Ideally zero. A nonzero value on this indicates that clients attempt to connect by using invalid credentials. Ensure that all clients are using valid credentials.

Consume message conversion time

Indicates that the accumulated time spent performing message conversion from clients that are consuming by using older protocol versions.

Table 8. Consume message conversion time metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_consume_conversions_time_quantile
Metric Type gauge
Value Type second
Segment By Service instance, Quantile, Service instance name

Ideally zero, as nonzero indicates that clients are experiencing more latency because of using an older protocol level. Those clients are down-level and must be upgraded. Ensure that all clients are at the latest levels.

Connected clients software name and version

The number of connected clients with a particular client software name and version.

Table 10. Connected clients software name and version metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_connected_clients_software_name_and_version
Metric Type gauge
Value Type number
Segment By Client software name, Client software version

This is for information to help you monitor the software name and version data of the active clients that are connected to the Event Streams instance.

Client software name and version are available for the Kafka client (Java version 2.4 or later, and other implementations that support software name and version) as described in KIP-8855. If the client software name and version are not available, these are set as unknown.

Inactive consumer groups

The number of inactive consumer groups in an Event Streams instance.

Table 11. Inactive consumer groups metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_inactive_consumergroups
Metric Type gauge
Value Type none
Segment By Service instance, Service instance name

This is for information only and is not an issue. Spikes indicate that a set of consumer groups stopped sending messages.

Instance bytes in per second

The number of bytes produced per second to an Event Streams instance.

Table 12. Instance bytes in per second metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_bytes_in_per_second
Metric Type gauge
Value Type byte
Segment By Service instance, Service instance name

This is for information to help you monitor trends in your usage of how many incoming or outgoing MB/s your clients are transferring to and from your cluster. Refer to Event Streams to determine what the recommended limits are for your plan and cluster.

Instance bytes out per second

The number of bytes consumed per second from an Event Streams instance.

Table 13. Instance bytes out per second metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_bytes_out_per_second
Metric Type gauge
Value Type byte
Segment By Service instance, Service instance name

This is for information to help you monitor trends in your usage of how many incoming or outgoing MB/s your clients are transferring to and from your cluster. Refer to Event Streams to determine what the recommended limits are for your plan and cluster.

Missing SNI connections

Incrementing count of the number of connections rejected due to not supporting the SNI extension to TLS.

Table 14. Missing SNI connections metric metadata
Metadata Description
Metric Name ibm_eventstreams_kafka_missing_sni_host_total
Metric Type counter
Value Type none
Segment By Service instance, Service instance name

Ideally this should be zero. It indicates clients that are not configured correctly. Clients must use the SNI extension for TLS to connect to the service. If this value is nonzero, ensure that all clients are at correct level and configured correctly for SNI.

Number of offline partitions

The number of partitions offline in an Event Streams instance.

Table 15. Number of offline partitions metric metadata
Metadata Description
Metric Name ibm_eventstreams_kafka_offline_partitions
Metric Type gauge
Value Type none
Segment By Service instance

Ideally this value should be zero. A nonzero value might indicate to a temporary issue with the cluster. It might also indicate to a Kafka partition leader election difficulty.

Number of partitions

The number of leader partitions in an Event Streams instance.

Table 16. Number of partitions metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_partitions
Metric Type gauge
Value Type none
Segment By Service instance, Service instance name

This is for information to help you monitor trends in your usage. Refer to Event Streams to determine what the recommended limits are for your plan and cluster.

Number of topics

The number of topics in an Event Streams instance.

Table 17. Number of topics metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_topics
Metric Type gauge
Value Type none
Segment By Service instance, Service instance name

Number of under in-sync replica partitions

The number of partitions with fewer than two in-sync replicas.

Table 18. Number of under in-sync replica partitions metric metadata
Metadata Description
Metric Name ibm_eventstreams_kafka_under_minisr_partitions
Metric Type gauge
Value Type none
Segment By Service instance

Ideally this value should be zero. A nonzero value might highlight a temporary issue with the cluster.

Produce message conversion time

Indicates that the accumulated time spent performing message conversion from clients that are producing by using older protocol versions.

Table 19. Produce message conversion time metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_produce_conversions_time_quantile
Metric Type gauge
Value Type second
Segment By Service instance, Quantile, Service instance name

Ideally zero. A consistent growth in this indicates that some clients are down-level and should be upgraded. Ensure that all clients are at the latest levels.

Rebalancing consumer groups

The number of rebalancing consumer groups in an Event Streams instance.

Table 20. Rebalancing consumer groups metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_rebalancing_consumergroups
Metric Type gauge
Value Type none
Segment By Service instance, Service instance name

While it is expected that this figure is occasionally >0 (as broker restarts happen frequently,) sustained high levels suggest that consumers might be restarting frequently and leaving or rejoining the consumer groups. Check you client logs.

Reserved disk space percentage

The percentage of reserved disk space that is required for all allocated partitions if fully used.

Table 21. Reserved disk space percentage metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_reserved_disk_space_percent
Metric Type gauge
Value Type percent
Segment By Service instance, Service instance name

Shows the percentage of disk space that would be used if your topics were filled to the extent of their configured retention size.

Schema greatest version percentage

The percentage of schema version capacity used for the schema with the greatest number of versions in the registry.

Table 22. Schema greatest version percentage metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_schema_registry_schema_versions_greatest_percentage
Metric Type gauge
Value Type percent
Segment By Service instance, Service instance name

Schema used percentage

The percentage of schema capacity used in the schema registry.

Table 23. Schema used percentage metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_schema_registry_schemas_used_percentage
Metric Type gauge
Value Type percent
Segment By Service instance, Service instance name

Stable consumer groups

The number of stable consumer groups in an Event Streams instance.

Table 24. Stable consumer groups metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_stable_consumergroups
Metric Type gauge
Value Type none
Segment By Service instance, Service instance name

Use along with rebalancing consumer groups. If this is consistently zero and rebalancing high, then it indicates a cluster problem. If this is nonzero and rebalancing high, it indicates a consumer group issue.

Topic bytes in per second

The number of bytes produced per second to a topic.

Table 25. Topic bytes in per second metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_topic_bytes_in_per_second
Metric Type gauge
Value Type byte
Segment By Service instance, IBM Event Streams Kafka topic, Service instance name

This is for information to help you monitor trends in your usage, particularly if any topics are producing unusual throughput, which is more or less than expected.

Topic bytes out per second

The number of bytes consumed per second from a topic.

Table 26. Topic bytes out per second metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_topic_bytes_out_per_second
Metric Type gauge
Value Type byte
Segment By Service instance, IBM Event Streams Kafka topic, Service instance name

This is for information to help you monitor trends in your usage, particularly if any topics are consuming unusually more or less throughput than expected.

IAM ID bytes in per second

The number of bytes in per second from an IAM ID.

Table 27. The number of bytes in per second per IAM ID
Metadata Description
Metric Name ibm_eventstreams_iam_id_bytes_in_per_second
Metric Type gauge
Value Type byte
Segment By Service instance, Service instance name, Iam Id

This is for information to help you monitor trends in your usage, particularly if any IAM IDs are producing unusually more throughput than expected.

This metric allows any differences in the amount of data being sent to the service from different users (IAM IDs) to be seen, and if required, guide the setting of any quotas needed.

IAM ID bytes out per second

The number of bytes out per second from an IAM ID.

Table 28. The number of bytes out per second per IAM ID
Metadata Description
Metric Name ibm_eventstreams_iam_id_bytes_out_per_second
Metric Type gauge
Value Type byte
Segment By Service instance, Service instance name, Iam Id

This is for information to help you monitor trends in your usage, particularly if any IAM IDs are consuming unusually more throughput than expected.

This metric allows you to see any differences in the amount of data that is sent from the service to different users (IAM IDs), and if required, guides the setting of any quotas needed.

IAM ID bytes in quota used percentage

The percentage of bytes in quota used per IAM ID. Where a bytes in quota was set for a user (IAM ID), this value shows the percentage of that quota being used.

Table 29. The percentage of bytes in quota used per IAM ID
Metadata Description
Metric Name ibm_eventstreams_iam_id_bytes_in_quota_used_percentage
Metric Type gauge
Value Type percent
Segment By Service instance, Service instance name, Iam Id

This is for information to help you monitor trends in your usage, particularly if any IAM IDs are producing close to their quota limits.

Quota metrics might sometimes exceed 100%. Kafka quotas use sampling and are applied asynchronously. For some workloads, especially where data is sent in large batches, this might result in small deviations from the limit.

IAM ID bytes out quota used percentage

The percentage of bytes out quota used per IAM ID. Where a bytes out quota was set for a user (IAM ID), this value shows the percentage of that quota being used.

Table 30. The percentage of bytes out quota used per IAM ID
Metadata Description
Metric Name ibm_eventstreams_iam_id_bytes_out_quota_used_percentage
Metric Type gauge
Value Type percent
Segment By Service instance, Service instance name, Iam Id

This is for information to help you monitor trends in your usage, particularly if any IAM IDs are consuming close to their quota limits.

Quota metrics might sometimes exceed 100%. Kafka quotas use sampling and are applied asynchronously. For some workloads, especially where data is sent in large batches, this might result in small deviations from the limit.

Used disk space percentage

The percentage of currently used disk space.

Table 31. Used disk space percentage metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_utilised_disk_space_percent
Metric Type gauge
Value Type percent
Segment By Service instance, Service instance name

This is for information to help you monitor trends in your usage. Refer to Event Streams to determine what the recommended limits are for your plan and cluster.

Rest-producer requests per second

Number of requests per second made to the REST Producer API.

Table 32: Rest-producer requests per second metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_rest_producer_requests_per_sec
Metric Type gauge
Value Type none
Segment By Service instance, Service instance name

This is for information to help you monitor usage of the REST Producer API, including use of schema encoders.

Mirroring throughput

The bytes per second of mirroring throughput from the source Event Streams instance.

Table 33. Mirroring throughput
Metadata Description
Metric Name ibm_eventstreams_instance_mirroring_throughput_bytes_per_second
Metric Type gauge
Value Type bytes_per_second
Segment By Service instance, IBM Event Streams Kafka topic, Service instance name

This is useful to see whether mirroring is active and for capacity planning.

Mirroring_latency

The per-topic mirroring latency in seconds from the source Event Streams instance.

Table 34. Mirroring latency
Metadata Description
Metric Name ibm_eventstreams_instance_mirroring_latency_seconds
Metric Type gauge
Value Type seconds
Segment By Service instance, IBM Event Streams Kafka topic, Service instance name

This is useful to determine how far behind a topic on the target cluster is.

Consumer group lag

Lag for each consumer group for each topic-partition in an Event Streams instance. This metric indicates that the number of messages that are yet to be processed for each partition in a consumer group.

Table 35. Consumer group lag metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_consumer_groups_lag
Metric Type gauge
Value Type none
Segment By Service instance name, IBM Event Streams consumer groups, IBM Event Streams Kafka topic, IBM Event Streams Kafka partitions

An increasing lag might highlight that the consumers in the group are not keeping pace with the rate that messages are being produced. This might require you to scale the number of consumers that process messages for the group.

It is normal for this metric to fluctuate when viewed over short time periods because of sampling and batch processing effects.

Message rate per partition

The rate of change of this metric gives the message per second that is incoming in to a partition of a Event Streams instance topic.

Table 36. Message rate per partition metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_message_rate_per_partition
Metric Type gauge
Value Type none
Segment By Service instance name, IBM Event Streams Kafka topic, IBM Event Streams Kafka partitions

Maximum partition retention percentage

Maximum percentage of the retention size used for partitions of a topic.

Table 37. Maximum partition retention percentage metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_max_partition_retention_percent
Metric Type gauge
Value Type percent
Segment By Service instance, IBM Event Streams Kafka topic, Service instance name

Topic size

Total disk size of all partitions of a topic.

Table 38. Topic size metric metadata
Metadata Description
Metric Name ibm_eventstreams_instance_topic_size
Metric Type gauge
Value Type byte
Segment By Service instance, IBM Event Streams Kafka topic, Service instance name

Instance utilization

Instance utilization. The level of utilization of an Event Streams instance. This is a numeric value between zero and two (inclusive):

  • 0 indicates that the workload being processed by this instance is within the capacity of the instance. More precisely, the utilization level is under 80%.
  • 1 indicates that the workload being processed by this instance is approaching the capacity limit for the instance. Review whether it is appropriate to scale the service instance. More precisely, the utilization level is over 80% and under 95%.
  • 2 indicates the workload being processed by this instance is at the capacity limit for the instance. As a result of this, messaging latency might increase. Review whether it is appropriate to scale the service instance. More precisely, the utilization level is over 95%.
Table 39. Instance utilization
Metadata Description
Metric Name ibm_eventstreams_instance_utilization
Metric Type gauge
Value Type int
Segment By Service instance, Service instance name

Attributes for segmentation

Global attributes

The following attributes are available for segmenting all of the listed metrics.

Table 40. Global attributes
Attribute Attribute name Attribute description
Cloud Type ibm_ctype The Cloud type is a value of public, dedicated, or local.
Location ibm_location The location of the monitored resource - this might be a region, data center or global.
Scope ibm_scope The scope is the account, organization, or space GUID associated with this metric.
Service name ibm_service_name Name of the service that generates this metric.
Service instance ibm_service_instance The service instance GUID identifies the instance that the metric is associated with.
Service instance name ibm_service_instance_name The service instance name provides the user-provided name of the service instance that isn't necessarily a unique value that depends on the name that is provided by the user.
Resource ibm_resource The resource that is measured by the service - typically an identifying name or GUID.
Resource Type ibm_resource_type The type of the resource that is measured by the service.
Resource group ibm_resource_group_name The resource group name where the service instance was created.

More attributes

The following attributes are available for segmenting one or more attributes. See the individual metrics for segmentation options.

Table 41. More attributes
Attribute Attribute name Attribute description
Client software name ibm_eventstreams_clientsoftwarename Client software name.
Client software version ibm_eventstreams_clientsoftwareversion Client software version.
IBM Event Streams Consumer Group ibm_eventstreams_consumergroup IBM Event Streams consumer group.
IBM Event Streams Kafka partition ibm_eventstreams_partition IBM Event Streams Kafka partition.
IBM Event Streams Kafka topic ibm_eventstreams_topic IBM Event Streams Kafka topic.
Quantile ibm_quantile The quantile represented when a metric supports segmenting by quantile.
Service instance ibm_service_instance The service instance segment identifies the instance that the metric is associated with.
Service instance name ibm_service_instance_name The service instance name provides the user-provided name of the service instance, which isn't necessarily a unique value that depends on the name that is provided by the user.

For more information about enabling platform metrics from the Event Streams dashboard and viewing metrics, see Monitoring Event Streams metrics.