The scheduler is configurable. Most likely, your scheduler is configured to use "MostRequestedPriority" (see #1), which tries to put all the pods pods on the smallest number of nodes (so you can shutdown the excess nodes). This priority function is used when running on a cloud provider, where you can lower your infrastructure costs by deleting unnecessary nodes/VMs).
#1 https://docs.openshift.com/container-platform/3.6/admin_ guide/scheduling/scheduler.html#other-priorities On Thu, May 24, 2018 at 1:00 AM, Lionel Orellana <[email protected]> wrote: > Hi All, > > We have 20 worker nodes all with the same labels (they all have the same > specs). Our pods don't have any node selectors so all nodes are available > to all pods. > > What we are seeing is the scheduler constantly placing pods on nodes that > are already heavily usitised (in terms of memory and/or cpu) while other > nodes have plenty of capacity. > > We have places resource request on some pods and they continue to be > placed on the busy nodes. > > How can we help the scheduler make better decisions? > > -bash-4.2$ oc version > oc v3.6.0+c4dd4cf > kubernetes v1.6.1+5115d708d7 > features: Basic-Auth GSSAPI Kerberos SPNEGO > > openshift v3.6.173.0.21 > kubernetes v1.6.1+5115d708d7 > > > > Thanks > > > > > > > _______________________________________________ > users mailing list > [email protected] > http://lists.openshift.redhat.com/openshiftmm/listinfo/users > >
_______________________________________________ users mailing list [email protected] http://lists.openshift.redhat.com/openshiftmm/listinfo/users
