IFS requires 128 GB of memory but only 25GB used

This is the way k8s works - it reserves memory per pod even if it is not used. It reserves memory to guarantee it’s available when/if needed.
You can i theory reduce the reserved memory, but then there is a risk pods will be evicted with out-of-memory as k8s don’t encourage memory swapping.
IFS set “memory limit” = “memory request” to ensure memory stability.

There is a parameter that can be used to set memory request lower than the memory limit.
global:
   memRatio: 50

This is not a supported parameter (not documented) by IFS and it can might cause OOM evictions. Don’t use it in any production environment unless you have 100% confidence you know what you are doing. And any OOB after setting that parameter can not be sent to IFS support.
 

Hi ​@MCIPSTEV , 

It may also be worth reviewing the IFS Cloud scale setting, as a higher scale may result in higher resource requirements for pods.

In my experience, reducing the scale helped some previously pending pods start. It may therefore be worth checking whether adjusting the scale could help in your environment as well.

Thanks for the feedback, with the help of IFS we have no more OOM errors, we have scale at 100.
But we want to use more memory because we have plenty. 

Most of the time we have 2 pods per function. Wich pods are heavily used, and how many pods can we use ?

Hi ​@MCIPSTEV, Since you have additional memory available, you could consider increasing the number of pods for heavily used components, such as OData or Reporting, to help distribute the workload. However, having more memory available doesn't necessarily mean you can allocate all of it to IFS Cloud pods, as you also need to consider CPU capacity and the resource requirements of each component.