Memory Limits Allocation

Issue: 
Case number
TS014283635

We are receiving alerts from eks_commhub_prod_east_2 cluster stating that the below mentioned nodes are breaching the Memory Limits Allocation. 

ip-10-45-30-219.us-east-2.compute.internal
ip-10-45-31-81.us-east-2.compute.internal 
ip-10-45-31-159.us-east-2.compute.internal 
ip-10-45-30-17.us-east-2.compute.internal 
ip-10-45-30-212.us-east-2.compute.internal
ip-10-45-31-154.us-east-2.compute.internal 

On checking the issue further with the DEVOPS team we figured out that the nodes are having enough memory to support the APIs running in it but in Instana we can see the health status of the nodes are pointing it to be "CRITICAL". 

We have set a threshold of 70% for warning and 80% for critical for these nodes in Instana. I have attached as a screenshot which would show the Memory and CPU usage as on today which is less than 70%. But strangely, in Instana the health seems to be pointing to critical even though everything is good with it's health as confirmed by the DEVOPS/AWS team. 

Solution that the IBM Technical Team provided: 
The warning is in reference to "Resource Limits" associated with the node. 

This is related to the QoS that Kubernetes provides and not a reference to the actual consumed memory on the host. This is calculated by summing of all of the pod limits on the node. When this sum exceeds the available capacity of the node the issue is emitted. There are 3 levels of QoS in Kubernetes:
* Guaranteed (request == limit)
* Burstable (request < limit)
* Best Effort (not specified)

"Currently it's not possible to adjust the level at which this value triggers the issue but can be raised as an AHA Feature Request".