terraformでautoscalingしているサーバのcloudwatch alarmを自動設定する

aws-scripts-mon

cloudwatchに登録した情報に対してterraformでalarmを設定していきます。

aws-scripts-monで--autoscalingをくっつけることでdimensionsにAutoScalingGroupNameをつけさせます。

mon-put-instance-data.pl --mem-util --disk-path=/ --disk-space-util --auto-scaling --from-cron

これをautoscalingで起動しているサーバにcronを登録します。

iamのpolicyにec2:DescribeTags権限がないとAutoscalingGroupNameが取れずにdimensionsに追加されないのに注意が必要です。

autoscaling

下記のようにすればcloudwatchを設定できます。aws_sns_topic.notify_to_slack.arnおよびaws_autoscaling_group.frontは作成済みという前提です。

resource "aws_cloudwatch_metric_alarm" "front_cpu" {
    alarm_name = "${aws_autoscaling_group.front.name} CPU Utilization"
    comparison_operator = "GreaterThanOrEqualToThreshold"
    evaluation_periods = "2"
    metric_name = "CPUUtilization"
    namespace = "AWS/EC2"
    dimensions { "AutoScalingGroupName"="${aws_autoscaling_group.front.name}" }
    statistic = "Average"
    period = "300"
    threshold = "80"
    alarm_description = "${aws_autoscaling_group.front.name} CPU Utilization"
    alarm_actions = [ "${aws_sns_topic.notify_to_slack.arn}" ]
    insufficient_data_actions = [ "${aws_sns_topic.notify_to_slack.arn}" ]
    ok_actions = [ "${aws_sns_topic.notify_to_slack.arn}" ]
}

resource "aws_cloudwatch_metric_alarm" "front_mem" {
    alarm_name = "${aws_autoscaling_group.front.name} Memory Utilization"
    comparison_operator = "GreaterThanOrEqualToThreshold"
    evaluation_periods = "2"
    metric_name = "MemoryUtilization"
    namespace = "System/Linux"
    dimensions { "AutoScalingGroupName"="${aws_autoscaling_group.front.name}" }
    statistic = "Average"
    period = "300"
    threshold = "90"
    alarm_description = "${aws_autoscaling_group.front.name} Memory Utilization"
    alarm_actions = [ "${aws_sns_topic.notify_to_slack.arn}" ]
    insufficient_data_actions = [ "${aws_sns_topic.notify_to_slack.arn}" ]
    ok_actions = [ "${aws_sns_topic.notify_to_slack.arn}" ]
}

instance

通常のインスタンスに対してはdimensionsとしてInstanceIdを使います。

aws_instance作成時にcountを使っているならaws_cloudwatch_metric_alarmでもcountを使うことで対応できます。

resource "aws_cloudwatch_metric_alarm" "elasticsearch_cpu" {
    alarm_name = "${element(aws_instance.elasticsearch.*.id, count.index)} CPU Utilization"
    comparison_operator = "GreaterThanOrEqualToThreshold"
    evaluation_periods = "2"
    metric_name = "CPUUtilization"
    namespace = "AWS/EC2"
    dimensions { "InstanceId"="${element(aws_instance.elasticsearch.*.id, count.index)}" }
    statistic = "Average"
    period = "300"
    threshold = "80"
    alarm_description = "${element(aws_instance.elasticsearch.*.id, count.index)} CPU Utilization"
    alarm_actions = [ "${aws_sns_topic.notify_to_slack.arn}" ]
    insufficient_data_actions = [ "${aws_sns_topic.notify_to_slack.arn}" ]
    ok_actions = [ "${aws_sns_topic.notify_to_slack.arn}" ]
    count = "${var.elasticsearch_server_count}"
}
resource "aws_cloudwatch_metric_alarm" "elasticsearch_mem" {
    alarm_name = "${element(aws_instance.elasticsearch.*.id, count.index)} Memory Usage"
    comparison_operator = "GreaterThanOrEqualToThreshold"
    evaluation_periods = "2"
    metric_name = "MemoryUtilization"
    namespace = "System/Linux"
    dimensions { "InstanceId"="${element(aws_instance.elasticsearch.*.id, count.index)}" }
    statistic = "Average"
    period = "300"
    threshold = "90"
    alarm_description = "${element(aws_instance.elasticsearch.*.id, count.index)} CPU Utilization"
    alarm_actions = [ "${aws_sns_topic.notify_to_slack.arn}" ]
    insufficient_data_actions = [ "${aws_sns_topic.notify_to_slack.arn}" ]
    ok_actions = [ "${aws_sns_topic.notify_to_slack.arn}" ]
    count = "${var.elasticsearch_server_count}"
}