Which of the following methods allow for the QUICKEST detection time?
Use Amazon CloudWatch metrics provided by Elastic Load Balancing to calculate average latency. Alarm and stop deployment when latency increases beyond the defined threshold.
Use AWS Lambda and Elastic Load Balancing access logs to detect average latency. Alarm and stop deployment when latency increases beyond the defined threshold.
Use AWS CodeDeploy’s MinimumHealthyHosts setting to define thresholds for rolling back deployments. If these thresholds are breached, roll back the deployment.
Use Metric Filters to parse application logs in Amazon CloudWatch Logs. Create a filter for latency. Alarm and stop deployment when latency increases beyond the defined threshold.
Explanations:
CloudWatch metrics from Elastic Load Balancing (ELB) provide real-time metrics, such as latency, with a very short detection time. Using CloudWatch alarms directly linked to the deployment pipeline allows the quickest detection and response time to latency increases.
While Lambda and ELB access logs can be used to detect latency, this method introduces a delay due to the need to process logs. Lambda functions aren’t as fast as CloudWatch metrics for immediate latency detection.
AWS CodeDeploy’s MinimumHealthyHosts setting is used for ensuring a minimum number of healthy instances during deployments but doesn’t specifically address latency thresholds. It’s more focused on instance health, not performance metrics like latency.
Metric filters in CloudWatch Logs can detect latency, but parsing application logs adds complexity and delay compared to using CloudWatch metrics from ELB. Logs aren’t as timely as metrics for latency monitoring and may increase detection time.
As far as I’m aware, the answer is:
Use Amazon CloudWatch metrics provided by Elastic Load Balancing to calculate average latency. Alarm and stop deployment when latency increases beyond the defined threshold.
I value that the answer is:
Use Amazon CloudWatch metrics provided by Elastic Load Balancing to calculate average latency. Alarm and stop deployment when latency increases beyond the defined threshold.