Monitoring rules reference

Monitoring rules are configured on a per-resource basis, with rules for the following resources:

Most monitoring rules contain a configurable field called Alert severity, which is the severity granted to an alert when its condition is triggered. Some rules have a fixed severity, as noted in the tables below. Monitoring views can be configured to only send alerts that meet or exceed a certain severity.

Rule componentDescriptionExample options
Alert severitySeverity of monitoring report conditionLow, Medium, High

Agent rules

RuleConfigurationConditions
Agent last heartbeat timeConfigurableSingle condition
Agent manager last heartbeat timeConfigurableSingle condition
Agent manager version stale timeConfigurableSingle condition
Agent version stale timeConfigurableSingle condition
High CPU utilizationConfigurableSingle condition
JVM heap usage is close to the limitConfigurableSingle condition
Low disk spaceConfigurableSingle condition
Time until earliest keystore certificate expiresConfigurableSingle condition
Time until earliest truststore certificate expiresConfigurableSingle condition
Queue sizeConfigurableSingle condition

Agent last heartbeat time

Alerts when the agent bootstrapper's last heartbeat is older than a set threshold.

Rule componentDescriptionExample options
If value is greater thanAmount of time elapsed since the last heartbeat received from the agent bootstrapper10 minutes

We recommend setting this monitor value to 10 minutes.

Agent manager last heartbeat time

Alerts when the agent manager's last heartbeat is older than a set threshold.

Rule componentDescriptionExample options
If value is greater thanAmount of time elapsed since the last heartbeat received from the agent manager10 minutes

Agent manager version stale time

Alerts when the agent bootstrapper version has not been upgraded since a set threshold.

Rule componentDescriptionExample options
If value is greater thanAmount of time elapsed since the agent manager has been on an old version10 days

We recommend setting this monitor value to 10 days.

Agent version stale time

Alerts when the agent version has not been upgraded since a set threshold.

Rule componentDescriptionExample options
If value is greater thanAmount of time elapsed since the agent has been on an older version10 days

We recommend setting this monitor value to 10 days.

High CPU utilization

Alerts when the agent CPU utilization exceeds a set threshold.

Rule componentDescriptionExample options
If value is greater thanPercentage of CPU utilization80

We recommend setting this monitor value to 80 (%).

JVM heap usage is close to the limit

Alerts when the JVM heap usage exceeds a set threshold.

Rule componentDescriptionExample options
If value is greater thanPercentage of JVM heap used / JVM heap available70

We recommend setting this monitor value to 70 (%).

Low disk space

Alerts when the available disk space drops below a set threshold.

Rule componentDescriptionExample options
If value is less thanAvailable disk space10GB

We recommend setting this monitor value to 10GB.

Time until earliest keystore certificate expires

Alerts when a certificate in the agent's keystore will expire within a set threshold.

Rule componentDescriptionExample options
If value is less thanAmount of time until a certificate expires10 days

** We recommend setting this monitor value to medium severity at less than 30 days and high severity at less than 10 days.**

Time until earliest truststore certificate expires

Alerts when a certificate in the agent's truststore will expire within a set threshold.

Rule componentDescriptionExample options
If value is less thanAmount of time until a certificate expires10 days

We recommend setting this monitor value to medium severity at less than 30 days and high severity at less than 10 days.

Queue size

Alerts when the number of jobs queued on an agent exceeds a set threshold.

Rule componentDescriptionExample options
If value is greater thanThe number of jobs in the agent's job queue70

We recommend setting this monitor value to 70 (jobs).

Schedule rules

RuleConfigurationConditions
Consecutive schedule failuresConfigurableSingle condition
Schedule durationConfigurableSingle condition

Consecutive schedule failures

Alerts when the number of consecutive schedule failures meets or exceeds a set threshold. This does not count schedule runs that result in a cancelled build.

Rule componentDescriptionExample options
If value is greater than or equal toThreshold of consecutive schedule failures1

The default behavior for this monitor is to alert with medium severity at one failure and high severity at three failures, though these thresholds are highly dependent on the frequency and stability of the schedules that are included in the monitoring rule's scope.

Schedule duration

Alerts when a schedule is running longer than a set threshold.

Rule componentDescriptionExample options
If value is greater than or equal toThe duration of the schedule2 hours
This monitor is typically used on highly critical schedules to quickly inform whether or not the schedule will complete in the expected time. Due to the variable nature of schedules, this monitor is often schedule-scoped.

A user-caused failure is a job failure that results from a problem with the configuration or input data, such as an invalid schema, a malformed row, or insufficient permissions. These failures are distinguished from transient or infrastructure-related failures, which are not surfaced by these alerts because they are typically resolved automatically by retries.

RuleConfigurationConditions
Changelog jobs failingConfigurableSingle condition
Merge changes job failingConfigurableSingle condition
Sync jobs failingFixed high severitySingle condition
Scroll job failing on pipelineConfigurableSingle condition
Sync propagation delayConfigurableSingle condition
Liveness: Time since last successful checkpointConfigurableSingle condition
Invalid stream records detectedFixed high severitySingle condition
Invalid direct write records detectedFixed high severitySingle condition

Changelog jobs failing

Alerts when the "changelog" job for the object or link is failing on either the active pipeline or the replacement pipeline. Only alerts on user-caused failures.

Rule componentDescriptionExample options
If value is greater than or equal toThreshold of consecutive user-caused changelog job failures1

The default behavior for this monitor is to alert with medium severity at one failure and high severity at three failures.

Merge changes job failing

Alerts when the "merge changes" job for the object or link is failing on either the active pipeline or the replacement pipeline. Only alerts on user-caused failures.

Rule componentDescriptionExample options
If value is greater than or equal toThreshold of consecutive user-caused merge job failures1

The default behavior for this monitor is to alert with medium severity at one failure and high severity at three failures.

Sync jobs failing

Alerts with high severity when a foreground sync job for the object or link fails terminally on either the active pipeline or the replacement pipeline. A terminal failure means that the job will not be retried, either because the failure is not retryable or because the job encountered too many unrecoverable errors. This rule is non-configurable.

Scroll job failing on pipeline

Alerts when the "scroll" job for the object or link's active or replacement pipeline is failing. Scroll jobs are responsible for streaming data from the backing datasource to the object databases.

Rule componentDescriptionExample options
If value is greater than or equal toThreshold of consecutive scroll job failures3

The default behavior for this monitor is to alert with low severity at one failure, medium severity at three failures, and high severity at seven failures, and these values are configurable.

Sync propagation delay

Alerts when a dataset backing the object has a transaction with a sync time that exceeds a set threshold.

Rule componentDescriptionExample options
If value is greater than or equal toThreshold of time taken to sync a transaction1 day

Alerts if a stream backing the object or link has not completed a checkpoint within the configured threshold. This rule detects streams that are not running and streams that are failing to checkpoint. The default threshold is 2 minutes at high severity.

Rule componentDescriptionExample options
If value is greater than or equal toThreshold of time elapsed since the last successful checkpoint2 minutes

Invalid stream records detected

Alerts when records in an input stream contain format violations. The scroll job ignores these records. This rule is non-configurable, alerting with high severity when the number of ignored rows is greater than or equal to one.

Invalid direct write records detected

Alerts when direct-write records are rejected because of validation failures. This rule is non-configurable, alerting with high severity when the number of rejected records is greater than or equal to one.

Streaming dataset rules

Derived stream monitors

RuleConfigurationConditions
Last checkpoint durationConfigurableSingle condition
Liveness: Time since last successful checkpointConfigurableSingle condition
Consecutive checkpoint failuresConfigurableSingle condition
Checkpoint trigger failure rateConfigurableSingle condition
Total lagConfigurableSingle condition
Total lag V2ConfigurableSingle condition
Sustained lagConfigurableSingle condition
Checkpoint sizeConfigurableSingle condition
Total throughputConfigurableSingle condition
Total throughput V2 (less than)ConfigurableSingle condition
Total throughput V2 (greater than)ConfigurableSingle condition
Job Manager CPU usageConfigurableSingle condition
Job Manager memory usageConfigurableSingle condition
Task Manager CPU usageConfigurableSingle condition
Task Manager memory usageConfigurableSingle condition

Last checkpoint duration

Alerts if the last checkpoint took more time than the configured threshold to complete.

Rule componentDescriptionExample options
If value is greater thanThreshold of time taken to checkpoint7 minutes

Liveness: Time since last successful checkpoint

Alerts if the stream has not completed a checkpoint since the configured threshold. The default threshold configuration is 5 minutes. This monitor encompasses streams that are not running as well as streams failing a checkpoint.

Rule componentDescriptionExample options
If value is greater than or equal toThreshold of time elapsed since last checkpoint5 minutes

Consecutive checkpoint failures

Alerts when the number of consecutive checkpoint failures meets or exceeds the configured threshold. The default high-severity threshold is five failures.

Rule componentDescriptionExample options
If value is greater than or equal toThreshold of consecutive checkpoint failures5

Checkpoint trigger failure rate

Alerts when the number of checkpoint trigger failures over a configured window exceeds the specified threshold.

Rule componentDescriptionExample options
If value is greater thanThreshold of checkpoint trigger failures0
Time windowThe time period to count checkpoint trigger failures in5 minutes

Total lag

Alerts if a stream's lag (total unprocessed upstream records) exceeds the set threshold.

Rule componentDescriptionExample options
If value is greater thanThreshold of unprocessed upstream records1000

This monitor indicates that streaming transforms are taking too long to run, or there is a problem with the streaming transforms infrastructure.

Total lag V2

Alerts when the total number of unprocessed records from streaming inputs meets or exceeds the configured threshold throughout a 5-minute window. The default high-severity threshold is 1,000 records.

Rule componentDescriptionExample options
If value is greater than or equal toThreshold of unprocessed records1,000

Sustained lag

Alerts when the total number of unprocessed records stays at or above the configured threshold for the entire configured window. You can configure a window from 2 minutes through 1 hour; the default is 5 minutes.

Rule componentDescriptionExample options
If value is greater than or equal toThreshold of unprocessed records1,000
Time windowTime during which lag must remain above the threshold5 minutes

Checkpoint size

Alerts when the latest checkpoint size exceeds the configured threshold.

Rule componentDescriptionExample options
If value is greater thanThreshold of checkpoint size, in bytes100 bytes

Total throughput

Alerts if a stream's throughput (records processed per checkpoints) falls below the set threshold.

Rule componentDescriptionExample options
If value is less thanThreshold of records processed per checkpoint100

This monitor indicates that streaming transforms are taking too long to run, or there is a problem with the streaming transforms infrastructure.

Total throughput V2 (less than)

Alerts when the number of records processed per checkpoint from streaming inputs stays below the configured threshold throughout a 5-minute window.

Rule componentDescriptionExample options
If value is less thanThreshold of records processed per checkpoint100

Total throughput V2 (greater than)

Alerts when the number of records processed per checkpoint from streaming inputs stays above the configured threshold throughout a 5-minute window.

Rule componentDescriptionExample options
If value is greater thanThreshold of records processed per checkpoint100

Job Manager CPU usage

Alerts when Job Manager CPU utilization exceeds the configured threshold.

Rule componentDescriptionExample options
If value is greater thanThreshold of Job Manager CPU utilization80%

Job Manager memory usage

Alerts when Job Manager memory utilization exceeds the configured threshold.

Rule componentDescriptionExample options
If value is greater thanThreshold of Job Manager memory utilization80%

Task Manager CPU usage

Alerts when CPU utilization for a Task Manager meets or exceeds the configured threshold throughout a 10-minute window.

Rule componentDescriptionExample options
If value is greater than or equal toThreshold of Task Manager CPU utilization80%

Task Manager memory usage

Alerts when memory utilization for a Task Manager meets or exceeds the configured threshold throughout a 10-minute window.

Rule componentDescriptionExample options
If value is greater than or equal toThreshold of Task Manager memory utilization80%

Ingest stream monitors

RuleConfigurationConditions
Records ingested over selected windowConfigurableSingle condition
Total throughput (less than or equal to)ConfigurableSingle condition
Total throughput (greater than or equal to)ConfigurableSingle condition
High source stalenessConfigurableSingle condition

Records ingested over selected window

Alerts if the number of records ingested into the raw stream's live view over the selected time window was less than or equal to the configured threshold. Available windows are 5 minutes, 30 minutes, 1 hour, 4 hours, and 1 day.

Rule componentDescriptionExample options
If value is less than or equal toThreshold of ingested records per unit time100

Total throughput (less than or equal to)

Alerts when the number of records processed per second stays at or below the configured threshold throughout a 5-minute window.

Rule componentDescriptionExample options
If value is less than or equal toThreshold of records processed per second100

Total throughput (greater than or equal to)

Alerts when the number of records processed per second stays at or above the configured threshold throughout a 5-minute window.

Rule componentDescriptionExample options
If value is greater than or equal toThreshold of records processed per second100

High source staleness

Alerts when source data staleness exceeds the configured threshold. This rule is available only when latency sampling is enabled and the data connection task supplies a source timestamp.

Rule componentDescriptionExample options
If value is greater thanThreshold of source data staleness20 seconds

Live deployment rules

RuleConfigurationConditions
Live deployment heartbeatConfigurableSingle condition

Live deployment heartbeat

Alerts when deployment has not emitted a heartbeat for more than one minute.

Rule componentDescriptionExample options
If value is greater than or equal toThreshold of time elapsed since last heartbeat1 minute

Time series sync rules

Points written by the time series sync over last 5 or 30 minutes

Alerts if the number of points written by the time series sync over the last 5 or 30 minute window was less than or equal to the configured threshold.

Rule componentDescriptionExample options
If value is less than or equal toThreshold of points written per unit time100

Time series stream liveness: Time since last successful checkpoint

Alerts if the streaming time series sync has not completed a checkpoint since the configured threshold. The default high-severity threshold is 2 minutes.

Rule componentDescriptionExample options
If value is greater than or equal toThreshold of time elapsed since the last successful checkpoint2 minutes

Dataset rules

RuleConfigurationConditions
Time since job last succeededConfigurableSingle condition

Time since job last succeeded

Alerts when a job on a dataset has not succeeded within a specified time threshold. Unlike the "Time since last updated" health check, the following conditions count as a passing status for the monitor:

  • The job succeeded, but the transaction was aborted
  • The job succeeded, but no new data was added
Rule componentDescriptionExample options
If value is greater than or equal toAmount of time elapsed since a job last succeeded1 day

We recommend setting this monitor value based on your dataset's expected update frequency. For daily updates, set it to 1 day.

Geotemporal observation rules

RuleConfigurationConditions
Geotemporal observations sent over last 5 or 30 minutesConfigurableSingle condition

Geotemporal observations sent over last 5 or 30 minutes

Alerts if the number of geotemporal observations sent over the last 5 or 30 minute window was less than or equal to the configured threshold.

Rule componentDescriptionExample options
If value is less than or equal toThreshold of geotemporal observations sent per unit time100

Automation rules

The following rules apply to both automations and time series streaming automations.

RuleConfigurationConditions
Automation has no new evaluationsConfigurableSingle condition
Automation has no new triggersConfigurableSingle condition
Automation has been disabled by the systemFixed high severitySingle condition
Automation had repeated execution failures in a windowConfigurableSingle condition
Automation had repeated evaluation failures in a windowConfigurableSingle condition
Automation had a high number of effect execution failures in a windowConfigurableSingle condition

Automation has no new evaluations

Alerts if there has been no new evaluation since the configured threshold. Use this rule to detect performance degradation in an automation that should have been evaluated but did not. This rule does not alert when the automation has not been triggered.

Rule componentDescriptionExample options
If value is greater than or equal toThreshold of time elapsed since last automation evaluation1 hour

Automation has no new triggers

Alerts if there have been no new monitor triggers within the configured threshold. Use this rule to detect when an automation is not being triggered as expected.

Rule componentDescriptionExample options
If value is greater than or equal toThreshold of time elapsed since last automation trigger1 day

Automation has been disabled by the system

Alerts if an automation was disabled by the system due to reaching limits or triggering cycles. This rule is non-configurable, alerting with high severity when the automation is disabled.

Automation had repeated execution failures in a window

Alerts when the number of failed automation executions in the window exceeds the configured threshold. Use this rule to surface automations that keep running and failing rather than ones the system has already disabled.

Rule componentDescriptionExample options
If value is greater thanThreshold of number of failed executions0
Time windowThe time period to count failed executions in1 hour

Automation had repeated evaluation failures in a window

Alerts when the number of failed automation evaluations in the window exceeds the configured threshold. Use this rule to catch automations whose trigger conditions fail to evaluate, separately from failures that occur during execution.

Rule componentDescriptionExample options
If value is greater thanThreshold of number of failed evaluations0
Time windowThe time period to count failed evaluations in1 hour

Automation had a high number of effect execution failures in a window

Alerts when the number of failed effect executions in the window exceeds the configured threshold. Effects are the downstream actions or notifications the automation runs. Use this rule to catch automations whose effects fail even when their triggers and evaluations succeed.

Rule componentDescriptionExample options
If value is greater thanThreshold of number of failed effect executions0
Time windowThe time period to count failed effect executions in1 hour

Function rules

Function executions can fail for a variety of reasons. For a full list of failure types, see function failure types.

RuleConfigurationConditionsFailure types included
Function duration p95ConfigurableSingle conditionNot applicable
Number of function failures in windowConfigurableSingle conditionAll
Number of user-facing function failures in windowConfigurableSingle conditionUser-facing only
Number of non-user-facing function failures in windowConfigurableSingle conditionAll except user-facing
Function failure rate in windowConfigurableComposite (AND)All
Non-user-facing function failure rate in windowConfigurableComposite (AND)All except user-facing

Composite rules combine a failure-percentage threshold with a minimum run-count threshold. The rule triggers only when both conditions reach a configured severity, which prevents alerts based on too little data.

Function duration p95

Alerts when the p95 function duration exceeds the specified thresholds. The p95 is measured over a sliding window of recent data.

Rule componentDescriptionExample options
If value is greater thanThreshold of duration10s

Number of function failures in window

Alerts when the total number of failed executions of a function in the given window exceeds a given threshold. This rule tracks all failure types, including both user-facing and non-user-facing errors.

Rule componentDescriptionExample options
If value is greater thanThreshold of number of failures0
Time windowThe time period to count failures in1 hour

Number of user-facing function failures in window

Alerts when the number of user-facing function failures over a given window exceeds the specified thresholds. This rule tracks only user-facing errors thrown by function code.

Rule componentDescriptionExample options
If value is greater thanThreshold of number of failures0
Time windowThe time period to count failures in1 hour

Number of non-user-facing function failures in window

Alerts when the number of function failures over a given window exceeds the specified thresholds, excluding user-facing errors thrown by function code. This rule is useful for monitoring infrastructure and system-level failures without noise from expected user input errors.

Rule componentDescriptionExample options
If value is greater thanThreshold of number of failures0
Time windowThe time period to count failures in1 hour

Function failure rate in window

Alerts when both the percentage of function runs that fail and the total number of runs exceed the specified thresholds over a given window. The total run count condition prevents the rule from triggering on low-traffic functions.

Rule componentDescriptionExample options
Failure percentage (%) is greater thanThreshold for the percentage of runs that fail, from 0 through 10010
Total run count is greater thanMinimum number of runs required before the rule triggers10
Time windowThe time period used to calculate both conditions5 minutes

Non-user-facing function failure rate in window

Alerts when both the percentage of function runs that fail with non-user-facing errors and the total number of runs exceed the specified thresholds over a given window. This rule excludes user-facing errors thrown by function code. The total run count condition prevents the rule from triggering on low-traffic functions.

Rule componentDescriptionExample options
Failure percentage (%) is greater thanThreshold for the percentage of runs that fail with non-user-facing errors, from 0 through 10010
Total run count is greater thanMinimum number of runs required before the rule triggers10
Time windowThe time period used to calculate both conditions5 minutes

Action rules

Action executions can fail for a variety of reasons. For a full list of failure types, see action failure types.

RuleConfigurationConditionsFailure types included
Action duration p95ConfigurableSingle conditionNot applicable
Number of action failures in windowConfigurableSingle conditionAll
Number of non-user-facing action failures in windowConfigurableSingle conditionAll except user-facing function failures

Action duration p95

Alerts when the p95 action duration exceeds the specified thresholds. The p95 is measured over a sliding window of recent data.

Rule componentDescriptionExample options
If value is greater thanThreshold of duration10s

Number of action failures in window

Alerts when the total number of failed executions of an action in the given window exceeds a given threshold. This rule tracks all failure types, including both user-facing and non-user-facing errors.

Rule componentDescriptionExample options
If value is greater thanThreshold of number of failures0
Time windowThe time period to count failures in1 hour

Number of non-user-facing action failures in window

Alerts when the number of action failures over a given window exceeds the specified thresholds, excluding failures caused by user-facing errors thrown by function-backed action code. This rule is useful for monitoring infrastructure and system-level failures without noise from expected user input errors.

This rule tracks all failure types except user-facing function failures thrown by function code and displayed to users.

Rule componentDescriptionExample options
If value is greater thanThreshold of number of failures0
Time windowThe time period to count failures in1 hour