)]}'
{"specs/kilo/instance-autorestart.rst":[{"author":{"_account_id":2284,"name":"Eoghan Glynn","email":"eglynn@redhat.com","username":"eglynn"},"change_message_id":"7f859319b01ecad191392ffcc733de1a2dbc2878","unresolved":false,"context_lines":[{"line_number":27,"context_line":""},{"line_number":28,"context_line":"These solutions are not fully working."},{"line_number":29,"context_line":"- OS::Heat::CWLiteAlarm is not working with OS::Ceilometer::Alarm when using an"},{"line_number":30,"context_line":"AutoscalingGroup."},{"line_number":31,"context_line":"More details can be found here:"},{"line_number":32,"context_line":"https://ask.openstack.org/en/question/47690/create-a-highly-available-auto-scaling-group/"},{"line_number":33,"context_line":"- OS::Ceilometer::Alarm with cpu_util \u003d\u003d 0 as a condition works only if an internal"}],"source_content_type":"text/x-rst","patch_set":2,"id":"5a890539_e1194589","line":30,"updated":"2014-11-13 10:59:56.000000000","message":"CWLiteAlarm is officially deprecated and should not be used going forward.\n\nFor one thing, it relies on running the Heat cfn-push-stats tool being run within the instance in order to report stats up the Heat CW API.\n\nWhereas the OS::Ceilometer::Alarm doesn\u0027t rely on anything special happening within the guest (though it does depend on the ceilometer compute agent running on the node, more on that point below)","commit_id":"5ea57740bdf37740d6d0263f0e97eb9e32999c43"},{"author":{"_account_id":4715,"name":"Angus Salkeld","email":"asalkeld@redhat.com","username":"asalkeld"},"change_message_id":"99ce51ae0ac603339b586deaa40b581e7b6b6829","unresolved":false,"context_lines":[{"line_number":27,"context_line":""},{"line_number":28,"context_line":"These solutions are not fully working."},{"line_number":29,"context_line":"- OS::Heat::CWLiteAlarm is not working with OS::Ceilometer::Alarm when using an"},{"line_number":30,"context_line":"AutoscalingGroup."},{"line_number":31,"context_line":"More details can be found here:"},{"line_number":32,"context_line":"https://ask.openstack.org/en/question/47690/create-a-highly-available-auto-scaling-group/"},{"line_number":33,"context_line":"- OS::Ceilometer::Alarm with cpu_util \u003d\u003d 0 as a condition works only if an internal"}],"source_content_type":"text/x-rst","patch_set":2,"id":"bab921fb_f401986b","line":30,"in_reply_to":"5a890539_e1194589","updated":"2015-04-24 00:05:33.000000000","message":"you also *can* send metrics from the guest\n\n1. by using requests and \"borrowing\" the credentials from os-collect-config\n\n2. by using cfn-push-stats (they get bounced from heat-api-cw to ceilometer - a bit ugly, but it works)\n\nIn your \"ask\", you said cfn-push-stats had an error, please report this as there *may* be a bug in : https://github.com/openstack/heat/blob/master/heat/engine/watchrule.py#L280-L298 (maybe things have changed in ceilometer client).","commit_id":"5ea57740bdf37740d6d0263f0e97eb9e32999c43"},{"author":{"_account_id":2284,"name":"Eoghan Glynn","email":"eglynn@redhat.com","username":"eglynn"},"change_message_id":"7f859319b01ecad191392ffcc733de1a2dbc2878","unresolved":false,"context_lines":[{"line_number":35,"context_line":"gets updated."},{"line_number":36,"context_line":"Using insufficient_data_action inside the alarm does not solve this problem since"},{"line_number":37,"context_line":"the states of the alarm are changing as followed:"},{"line_number":38,"context_line":"alarm/ok -\u003e insufficient_data -\u003e alarm/ok -\u003e insufficient_data etc."},{"line_number":39,"context_line":"That will cause an auto restart even if the instance is up and available."},{"line_number":40,"context_line":"I couldn\u0027t find a reason, but when using a binary meter (sample values are 1 or 0)"},{"line_number":41,"context_line":"this problem does not happen."}],"source_content_type":"text/x-rst","patch_set":2,"id":"5a890539_1c431815","line":38,"updated":"2014-11-13 10:59:56.000000000","message":"This flapping behaviour is likely to be caused by a mismatch between the interval defined for the cpu_source in your /etc/ceilometer/pipeline.yaml on the compute node (out of the box default is 600s) versus the period defined for the alarm (60s).\n\nThe period of the alarm must be a whole number multiple of the corresponding metric gathering period, in this case 60s.\n\nCan you try lining up these periods, and then restarting the compute agent on that node?\n\nOnce that flapping in and out of insufficient_data is addressed, then the alarm action could be based of a transition to insufficient_data indicating that the stream of metrics for that instance has really dried up (due to the entire node going down, or just the individual instance).","commit_id":"5ea57740bdf37740d6d0263f0e97eb9e32999c43"},{"author":{"_account_id":4715,"name":"Angus Salkeld","email":"asalkeld@redhat.com","username":"asalkeld"},"change_message_id":"99ce51ae0ac603339b586deaa40b581e7b6b6829","unresolved":false,"context_lines":[{"line_number":35,"context_line":"gets updated."},{"line_number":36,"context_line":"Using insufficient_data_action inside the alarm does not solve this problem since"},{"line_number":37,"context_line":"the states of the alarm are changing as followed:"},{"line_number":38,"context_line":"alarm/ok -\u003e insufficient_data -\u003e alarm/ok -\u003e insufficient_data etc."},{"line_number":39,"context_line":"That will cause an auto restart even if the instance is up and available."},{"line_number":40,"context_line":"I couldn\u0027t find a reason, but when using a binary meter (sample values are 1 or 0)"},{"line_number":41,"context_line":"this problem does not happen."}],"source_content_type":"text/x-rst","patch_set":2,"id":"bab921fb_d496944b","line":38,"in_reply_to":"5a890539_1c431815","updated":"2015-04-24 00:05:33.000000000","message":"also the cooldown in your template is the same as the period. try increasing this to help with excessive actions.","commit_id":"5ea57740bdf37740d6d0263f0e97eb9e32999c43"},{"author":{"_account_id":4715,"name":"Angus Salkeld","email":"asalkeld@redhat.com","username":"asalkeld"},"change_message_id":"99ce51ae0ac603339b586deaa40b581e7b6b6829","unresolved":false,"context_lines":[{"line_number":49,"context_line":"This method will use the domain state in order to determine the instance health."},{"line_number":50,"context_line":""},{"line_number":51,"context_line":"The meter will return a binary value: 1 if the instance state is VIR_DOMAIN_RUNNING"},{"line_number":52,"context_line":"and if not it will return 0."},{"line_number":53,"context_line":""},{"line_number":54,"context_line":"Metric Definition:"},{"line_number":55,"context_line":" \u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\t\t\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\t\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\t\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d\u003d"}],"source_content_type":"text/x-rst","patch_set":2,"id":"bab921fb_b4f44087","line":52,"updated":"2015-04-24 00:05:33.000000000","message":"note this is a weaker test then the in guest heartbeat.","commit_id":"5ea57740bdf37740d6d0263f0e97eb9e32999c43"}]}