Compute Optimizer is no longer able to read CloudWatch memory metrics ?
I have installed CloudWatch agent in several EC2s in several accounts in different regions but from a more than a month ago Compute Optimizer is unable to read the Memory Metrics, has AWS changed something on the permissions?
The CloudWatch agent is generating the metrics:
Am I missing something ? Kind regards.
- Language
- English
- Newest
- Most votes
- Most comments
Hello @MVR-060709,
Thanks for your input. I have checked again the configuration and nothing has change from my side:
And as I said before, is happening in several accounts accros several regions.
Kind regards.
Were you able to solve this? I have same exact issue...
Hello @Karolis, it fixed by itself. Always check that the metrics are being shown in CloudWatch Metrics, if they are not there the issue is with CloudWatch Agent or the Instance Profile; if they are being shown but Compute Optimizer is not finding them check the NameSpace and metric's name to see if you are using the default names otherwise Compute Optimizer won't find them.
Hello! There are a number of different permissions which are required for Compute Optimizer to access the metrics and then also for those recommendation to be viewable to you. To view EC2 metrics in Compute Optimizer, the user/role you are logged in with would need the ec2:DescribeInstances permission. To view CloudWatch metrics in Compute Optimizer, you would need the cloudwatch:GetMetricData permission. (Source: https://docs.aws.amazon.com/compute-optimizer/latest/ug/security-iam.html)
From a Compute Optimizer perspective, I recommend you leverage a service-linked role (if you aren't already). This will ensure that Compute Optimizer has all of the permissions it requires to interact with other services. I've included a link to how to recreate the role if it has been inadvertently removed. https://docs.aws.amazon.com/compute-optimizer/latest/ug/using-service-linked-roles.html
Hope this is helpful!
Relevant content
- asked 5 years ago
It is now showing the Memory metric, definitely it was something inside AWS.