As workload is shifting from grid engine to kubernetes, we should downscale one and upscale the other.
As of this writing , a quick check at https://sge-status.toolforge.org/ shows that:
- there are 2x exec nodes running less than 8 jobs
- there are 3x weblight nodes running just 2 jobs
So perhaps we can just take that amount down and relocate the CPU/RAM into kubernetes.
@taavi has volunteered to do this work.