| { |
| "AMBARI": { |
| "service": [ |
| ], |
| "AMBARI_SERVER" : [ |
| { |
| "name": "ambari_server_agent_heartbeat", |
| "label": "Ambari Agent Heartbeat", |
| "description": "This alert is triggered if the server has lost contact with an agent.", |
| "interval": 2, |
| "help_url": "https://cwiki.apache.org/confluence/display/AMBARI/Ambari+Alerts#AmbariAlerts-ambari_agent_heartbeat", |
| "scope": "HOST", |
| "enabled": true, |
| "source": { |
| "type": "SERVER", |
| "class": "org.apache.ambari.server.alerts.AgentHeartbeatAlertRunnable" |
| } |
| }, |
| { |
| "name": "ambari_server_stale_alerts", |
| "label": "Ambari Server Alerts", |
| "description": "This alert is triggered if the server detects that there are alerts which have not run in a timely manner.", |
| "interval": 5, |
| "scope": "SERVICE", |
| "enabled": true, |
| "source": { |
| "type": "SERVER", |
| "class": "org.apache.ambari.server.alerts.StaleAlertRunnable", |
| "parameters": [ |
| { |
| "name": "stale.interval.multiplier", |
| "display_name": "Interval Multiplier", |
| "value": 2, |
| "type": "NUMERIC", |
| "description": "The number of intervals which must pass before an an alert is considered stale. This value is a multiplier for each alert's individual interval.", |
| "units": "Intervals", |
| "threshold": "CRITICAL" |
| } |
| ] |
| } |
| }, |
| { |
| "name": "ambari_server_performance", |
| "label": "Ambari Server Performance", |
| "description": "This alert is triggered if the Ambari Server detects that there is a potential performance problem with Ambari. This type of issue can arise for many reasons, but is typically attributed to slow database queries and host resource exhaustion.", |
| "interval": 5, |
| "scope": "SERVICE", |
| "enabled": true, |
| "source": { |
| "type": "SERVER", |
| "class": "org.apache.ambari.server.alerts.AmbariPerformanceRunnable", |
| "parameters": [ |
| { |
| "name": "request.by.status.warning.threshold", |
| "display_name": "Get Request Progress", |
| "value": 3000, |
| "type": "NUMERIC", |
| "description": "The time to find requests in progress before a warning alert is triggered.", |
| "units": "ms", |
| "threshold": "WARNING" |
| }, |
| { |
| "name": "request.by.status.critical.threshold", |
| "display_name": "Get Request Progress", |
| "value": 5000, |
| "type": "NUMERIC", |
| "description": "The time to find requests in progress before a critical alert is triggered.", |
| "units": "ms", |
| "threshold": "CRITICAL" |
| }, |
| { |
| "name": "task.status.aggregation.warning.threshold", |
| "display_name": "Get Request Status", |
| "value": 3000, |
| "type": "NUMERIC", |
| "description": "The time to calculate a request's status from its tasks before a warning alert is triggered.", |
| "units": "ms", |
| "threshold": "WARNING" |
| }, |
| { |
| "name": "task.status.aggregation.critical.threshold", |
| "display_name": "Get Request Status", |
| "value": 5000, |
| "type": "NUMERIC", |
| "description": "The time to calculate a request's status from its tasks before a critical alert is triggered.", |
| "units": "ms", |
| "threshold": "CRITICAL" |
| }, |
| { |
| "name": "rest.api.cluster.warning.threshold", |
| "display_name": "Get Cluster", |
| "value": 5000, |
| "type": "NUMERIC", |
| "description": "The time to get a cluster via the REST API before a warning alert is triggered.", |
| "units": "ms", |
| "threshold": "WARNING" |
| }, |
| { |
| "name": "rest.api.cluster.critical.threshold", |
| "display_name": "Get Cluster", |
| "value": 7000, |
| "type": "NUMERIC", |
| "description": "The time to get a cluster via the REST API before a critical alert is triggered.", |
| "units": "ms", |
| "threshold": "CRITICAL" |
| } |
| ] |
| } |
| }, |
| { |
| "name": "ambari_server_component_version", |
| "label": "Component Version", |
| "description": "This alert is triggered if the server detects that there is a problem with the expected and reported version of a component. The alert is suppressed automatically during an upgrade.", |
| "interval": 5, |
| "scope": "SERVICE", |
| "enabled": true, |
| "source": { |
| "type": "SERVER", |
| "class": "org.apache.ambari.server.alerts.ComponentVersionAlertRunnable" |
| } |
| } |
| ], |
| "AMBARI_AGENT" : [ |
| { |
| "name": "ambari_agent_disk_usage", |
| "label": "Host Disk Usage", |
| "description": "This host-level alert is triggered if the amount of disk space used goes above specific thresholds. The default threshold values are 50% for WARNING and 80% for CRITICAL.", |
| "interval": 1, |
| "scope": "HOST", |
| "enabled": true, |
| "source": { |
| "type": "SCRIPT", |
| "path": "alert_disk_space.py", |
| "parameters": [ |
| { |
| "name": "minimum.free.space", |
| "display_name": "Minimum Free Space", |
| "value": 5000000000, |
| "type": "NUMERIC", |
| "description": "The overall amount of free disk space left before an alert is triggered.", |
| "units": "bytes", |
| "threshold": "WARNING" |
| }, |
| { |
| "name": "percent.used.space.warning.threshold", |
| "display_name": "Warning", |
| "value": 50, |
| "type": "PERCENT", |
| "description": "The percent of disk space consumed before a warning is triggered.", |
| "units": "%", |
| "threshold": "WARNING" |
| }, |
| { |
| "name": "percent.free.space.critical.threshold", |
| "display_name": "Critical", |
| "value": 80, |
| "type": "PERCENT", |
| "description": "The percent of disk space consumed before a critical alert is triggered.", |
| "units": "%", |
| "threshold": "CRITICAL" |
| } |
| ] |
| } |
| }, |
| { |
| "name": "ambari_agent_ulimit", |
| "label": "Ulimit for open files", |
| "description": "This host-level alert is triggered if value of ulimit for open files (-n) goes above specific thresholds. The default threshold values are 200000 for WARNING and 800000 for CRITICAL.", |
| "interval": 1, |
| "scope": "HOST", |
| "enabled": true, |
| "source": { |
| "type": "SCRIPT", |
| "path": "alert_ulimit.py", |
| "parameters": [ |
| { |
| "name": "ulimit.warning.threshold", |
| "display_name": "Warning", |
| "value": 200000, |
| "type": "NUMERIC", |
| "description": "The threshold of ulimit for open files (-n) for WARNING alert.", |
| "threshold": "WARNING" |
| }, |
| { |
| "name": "ulimit.critical.threshold", |
| "display_name": "Critical", |
| "value": 800000, |
| "type": "NUMERIC", |
| "description": "The threshold of ulimit for open files (-n) for CRITICAL alert.", |
| "threshold": "CRITICAL" |
| } |
| ] |
| } |
| }, |
| { |
| "name": "ambari_agent_version_select", |
| "label": "Ambari Agent Distro/Conf Select Versions", |
| "description": "This host-level alert is triggered if the distro selector such as hdp-select cannot calculate versions available on this host. This may indicate that /usr/$stack/ directory has links/dirs that do not belong inside of it.", |
| "interval": 5, |
| "scope": "HOST", |
| "enabled": true, |
| "source": { |
| "type": "SCRIPT", |
| "path": "alert_version_select.py" |
| } |
| } |
| |
| ] |
| } |
| } |
