Hi Benoit, Two observations: you may need to put PIG as the component name, and the script may have to be a Python file. You can try changing both of these things and see if it starts functioning, then work back to figure out what isn't supported.
You can certainly add checks not related to a service, since some run on at the Host-level. I believe the metric source is the type, such as a port, script, url, etc. Thanks, Alejandro From: Benoit Perroud <[email protected]<mailto:[email protected]>> Reply-To: "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>> Date: Wednesday, December 30, 2015 at 4:57 AM To: "[email protected]<mailto:[email protected]>" <[email protected]<mailto:[email protected]>> Subject: Custom Alert definition Hi All, I'm struggling with the following use case: I want to add a custom alert which should call a custom script monitoring some Pig jobs. My alert definition currently looks like this: "AlertDefinition" : { "cluster_name" : "MyCluster", "component_name" : null, "description" : "My description", "enabled" : true, "ignore_host" : true, "interval" : 5, "label" : "My_Label", "name" : "my_name", "scope" : "ANY", "service_name" : "PIG", "source" : { "parameters" : [ { "name" : "connection.timeout1", "display_name" : "Connection Timeout1", "units" : "seconds", "value" : 10.0, "description" : "The maximum time before this alert is considered to be CRITICAL", "type" : "NUMERIC", "threshold" : "CRITICAL" } ], "path" : "/usr/local/bin/pig_check.sh", "type" : "SCRIPT" } } My alert is added properly, i.e. I'm able to query api/v1/clusters/DAPLAB02/alert_definitions/${alert_id}, but I don't see my script called at all, i mean, the state in Ambari is NONE, and I don't see any alert (api/v1/clusters/DAPLAB02/alerts) with defintion_id = ${alert_id}. Also, is there some more detailed documentation about the properties such as - ignore_host - what is the difference between name and description? - can we add general checks not related to any service? - with the METRIC source, can we monitor metrics reported in the Ambari Collector instead of JMX? Thanks in advance, Benoit.
