Posted On: Dec 8, 2017
We are excited to announce that Amazon CloudWatch Alarms now supports alarming when any M out of N datapoints cross the alarm threshold. A datapoint is the value of a metric for a given metric aggregation period i.e. if you use one minute as an aggregation period for a metric, then there will be one datapoint every minute. With this launch, you can now create a CloudWatch alarm that alerts you when M out of N datapoints of a metric are above your predefined threshold, such as three out of five times in any given five minutes interval or two out of six times in a thirty minutes interval. When any M out of N datapoints are below your threshold in an interval, the alarm will be in OK state. Please note that the M datapoints out of N datapoints in an interval can be of any order and does not need to be consecutive. Consequently, you can now get alerted even when the spikes in your metrics are intermittent over an interval.
This feature gives you the flexibility to configure CloudWatch alarms in ways that fit your use cases. You can create alarms that alerts you even if one datapoint is above your threshold or only when all the datapoints in an interval have breached the threshold or any configuration between those use cases. To create alarms with this feature, please visit Amazon CloudWatch console. The PutMetricAlarm API has also been updated with a new a parameter, DatapointsToAlarm, that defines the number of datapoints that must be breaching to trigger the alarm. There is no additional cost to use this feature.