Cruise Control dashboard in SMM UI

Cruise Control Dashboard

You can manually refresh the Dashboard using , or you can set a refresh interval by selecting one of the options available with .

Execution

The Execution tile of the Dashboard allows you to track and monitor the tasks being executed by Cruise Control (including rebalancing tasks that are triggered by the enabled goals or other operations started through the REST API). If the goals are fulfilled and there is no ongoing execution, the No task in progress message is shown. Only users with Admin level can stop the execution process.

Monitor

The Monitor tile always shows the ratio of Valid Windows, Valid Partitions, and Coverage of the Kafka cluster.

Valid windows

The metrics that Cruise Control collects about brokers and partitions in a Kafka cluster are aggregated in windows of time. A window represents a time range. There is a minimum number of metric samples that a window should have to be considered "valid" by Cruise Control. Only valid windows can be used for balance proposals and anomaly detection.

The validity of a window is determined based on valid brokers, valid partitions and overall valid metric samples count. The parameters of this mechanism can be configured in Cloudera Manager with the following Cruise Control configuration properties:

Property	Description
num.broker.metrics.windows	The maximum number of broker windows the Load Monitor keeps. Each window covers a time window defined by the broker.metrics.window.ms configuration property.
broker.metrics.window.ms	The size of the window in milliseconds to aggregate the Kafka broker metrics. The window must be greater than the value of the metric.sampling.interval.ms configuration property. metric.sampling.interval.ms is the duration until metrics are collected and sampled. When an interval is ready, another interval starts and tries to collect the metrics from the brokers.
min.samples.per.broker.metrics.window	The minimum number of metric samples a valid broker window should have. If a broker does not have enough samples in a broker window, the broker is removed from the window due to insufficient data.
max.allowed.extrapolations.per.broker	The maximum number of extrapolations allowed for each broker. A broker is considered as invalid if the total number extrapolations in all the windows goes above the maximum allowed number. Load Monitor contains information about the status of the linear regression model, which is used to estimate the CPU utilization, number of valid windows and partitions. If there are flawed brokers, not all windows have metrics. Cruise Control extrapolates these metric values.

Valid partitions

Cruise Control aggregates metrics about each partition in the connected Kafka cluster. There is a minimum number of metric samples a valid partition should have. The validity of a partition is determined based on the valid metric samples count. The mechanism can be configured in Cloudera Manager with the following Cruise Control configuration properties:

Property	Description
num.partition.metrics.windows	The maximum number of partition windows the Load Monitor keeps. Each window covers a time window defined by the partition.metrics.window.ms configuration property.
partition.metrics.window.ms	The size of the window in milliseconds to aggregate the Kafka partition metrics. The window must be greater than the value of the metric.sampling.interval.ms configuration property. metric.sampling.interval.ms is the duration until metrics are collected and sampled. When an interval is ready, another interval starts and tries to collect the metrics from the brokers.
min.samples.per.partition.metrics.window	The minimum number of metric samples a valid partition window should have. If a partition does not have enough samples in a partition window, the topic of the partition is removed from the window due to insufficient data.
max.allowed.extrapolations.per.partition	The maximum number of extrapolations allowed for each partition. A partition is considered as invalid if the total number extrapolations in all the windows goes above this number. Load Monitor contains information about the status of the linear regression model, which is used to estimate the CPU utilization, number of valid windows and partitions. If there are flawed partitions, not all windows have metrics. Cruise Control extrapolates these metric values.

Coverage

The coverage of this cluster model through monitored partitions ratio, showing how representative the cluster is.

List of Goals

The Goals tile shows the list of enabled goals and indicates if the goals have been fulfilled or not with checkmarks. The list of goals can be configured in Cloudera Manager. For more information, see Setting capacity estimates and goals.

List of Anomaly Detectors

The Anomaly Detector tile shows the list of anomaly detectors and whether self-healing is enabled for the given detector. If there are any anomalies found by the detectors, the anomalies are listed under the corresponding detector.

Cluster State Dashboard

The Cluster State Dashboard provides an overview of the Kafka cluster specific metrics related to leaders, replicas, and directories. The metrics shown on the Dashboard are based on the number of leaders, online, offline and out-of-sync replicas, online and offline directories within the brokers available on the Kafka cluster. You can use these metrics to monitor the Kafka cluster state.

In the Summary section, you can review the general information of the number of brokers, topics, replicas and leaders within the Kafka cluster, and more metrics that show the average, maximum and standard ratio of leaders per broker and replicas per broker.