Metrics parameters (cdap-site.xml and cdap.default.xml)

metrics.bind.address

0.0.0.0

Metrics query HTTP service bind address.

metrics.bind.port

0

Metrics query HTTP service bind port; if 0, binds to a random port.

metrics.boss.threads

${http.service.boss.threads}

Number of Netty service boss threads for metrics HTTP services.

metrics.connection.backlog

${http.service.connection.backlog}

Maximum connection backlog of metrics HTTP service.

metrics.minimum.resolution.seconds

1

Minimum metrics resolution in seconds, this value should be less than 60 seconds, otherwise the minimum metrics resolution will be 60 seconds.

metrics.data.table.retention.minimum.resolution.seconds

7200

Retention resolution in seconds of the minimum resolution table; default retention period is 2 hours, this retention will not be applied if the minimum resolution is greater than or equal to 60.

metrics.data.table.retention.resolution.3600.seconds

2592000

Retention resolution in seconds of the 1-hour resolution table; default retention period is 30 days.

metrics.data.table.retention.resolution.60.seconds

2592000

Retention resolution in seconds for the 1-minute resolution table; default retention period is 30 days.

metrics.data.table.ts.rollTime.3600

24

Number of columns in a 1-hour resolution timeseries table.

metrics.data.table.ts.rollTime.60

60

Number of columns in a 1-minute resolution timeseries table.

metrics.dataset.hbase.stats.report.interval

60

Report interval in seconds for HBase stats.

metrics.dataset.leveldb.stats.report.interval

60

Report interval in seconds for LevelDB stats.

metrics.exec.threads

${http.service.exec.threads}

Number of Netty service executor threads for metrics HTTP services.

metrics.meta.table

metrics.meta

Name of the metrics meta table.

metrics.kafka.partition.size

10

Number of partitions for the Kafka metrics topic.

metrics.kafka.topic.prefix

metrics

Topic prefix used to publish metrics in Kafka.

metrics.max.instances

${master.service.max.instances}

Maximum number of instances for the metrics service.

metrics.memory.mb

${master.service.memory.mb}

Memory in megabytes for each metrics service instance.

metrics.messaging.meta.table

metrics.messaging.meta

Name of the messaging metrics meta table.

metrics.messaging.topic.num

10

Number of topics for metrics messages. This property also sets the number of threads used to fetch and process metrics in parallel from the messaging service. For a value of N, topics will be created for metrics with names beginning at ${metrics.topic.prefix}0, ${metrics.topic.prefix}1, up to ${metrics.topic.prefix}(N-1).

metrics.num.cores

${master.service.num.cores}

Number of virtual cores for the metrics service.

metrics.num.instances

1

Number of instances for the metrics service.

metrics.processor.max.instances

${master.service.max.instances}

Maximum number of instances for metrics processor service Apache Twill runnable.

metrics.processor.memory.mb

${master.service.memory.mb}

Memory in megabytes for each metrics processor service Apache Twill runnable instance.

metrics.processor.num.cores

1

Number of virtual cores for metrics processor service Apache Twill runnable.

metrics.processor.num.instances

1

Number of instances for metrics processor service Apache Twill runnable.

metrics.processor.status.bind.address

0.0.0.0

Metrics processor HTTP service bind address.

metrics.processor.status.bind.port

0

Metrics processor HTTP service bind port; if 0, binds to a random port.

metrics.topic.prefix

metrics

Topic prefix used to publish metrics in messaging.

metrics.admin.topic

metricsadmin

Topic name for publishing metrics administrative events to the messaging system.

metrics.admin.poll.delay.millis

2000

The delay in milliseconds to check again for new metrics administrative events after it detects there was no event.

metrics.worker.threads

${http.service.worker.threads}

Number of Netty service worker threads for metrics HTTP services.

metrics.processor.queue.size

1000

Maximum size of a queue where the metrics processor temporarily stores newly-fetched metrics in-memory before persisting them.

metrics.processor.max.delay.ms

3000

Maximum delay in milliseconds allowed between the latest metrics timestamp and the time when it is processed.

app.program.metrics.enabled

true

Enable or disable emitting metrics from user application programs, by default set to true.

metrics.hbase.max.scan.threads

96

Maximum number of threads used for scanning HBase tables.

metrics.table.splits

16

Number of splits for all metrics tables. This property can only be changed before CDAP starts up for the first time and creates the metrics tables. Once all metrics tables are created, this property will not take any effect.

metrics.table.hbase.split.policy

org.apache.hadoop.hbase.regionserver.DisabledRegionSplitPolicy

The class name that controls the HBase table region split policy. Ideally, auto-splitting should be disabled for HBase tables used by the metrics system.

structured.table.time.metrics.enabled

false

Option to turn on the time metrics for the structured table operations.