Hello, i realized that hpa uses cpu/requests as base to calc pod's cpu consumption. We are overcommitting our nodes, and the value of 'targetPercentage' isn't the best fit, as a workaround i made the following calc:
L = 731 (cpu limit) R = 18 (cpu request) P = 80 (scale percentage) ((L * (P / 100))/R) * 100 So, for a pod with 18 milicores cpu/request and 731 milicores cpu/limit the correct value would be around 3248%. I've patched a hpa and it works fine, my question is why are we using cpu/request as base for cpu consumption and if there's any chance to use cpu/limit.
_______________________________________________ dev mailing list [email protected] http://lists.openshift.redhat.com/openshiftmm/listinfo/dev
