You should also be able to edit the `prometheus` api object created, for example to have a spec like:
> apiVersion: "monitoring.coreos.com/v1alpha1" > kind: "Prometheus" > metadata: > name: k8s > labels: > prometheus: k8s > spec: > retention: "2h" > resources: > requests: > memory: 500Mi Note that the actual memory usage needed will vary based on the services monitored, retention period, number of node-exporters running, etc etc. Setting the 'resources' field of the 'Prometheus' kind is a good way to change the resources of the created statefulset. - Euan On 05/24/2017 04:55 PM, Arve Knudsen wrote: > Aha thanks for letting me know. > > On Wed, May 24, 2017 at 8:41 PM Rob Szumski <[email protected] > <mailto:[email protected]>> wrote: > > This is not currently configurable at install time, but you should > be able to modify the object on the cluster that you already have > running. Change the limit parameter from 2GB down to like 500MB or > something, and then update -storage.local.target-heap-size to be 50% > of that. > > Once you save, the pod should get scheduled. > > For anyone following along, note that these types of changes are not > generally recommended/supported and there are no guarantees that > they will preserved upon a software update. > > - Rob > >> On May 24, 2017, at 11:01 AM, Arve Knudsen <[email protected] >> <mailto:[email protected]>> wrote: >> >> Thanks for letting me know, Rob. I'm bringing this cluster up for >> testing though, so I don't think you'll need much history. How can >> I configure Promotheus to use a certain amount of memory? >> >> Thanks, >> Arve >> >> On Wed, May 24, 2017 at 7:36 PM Rob Szumski >> <[email protected] <mailto:[email protected]>> wrote: >> >> Its default limit is 2GB, which allows it to keep a decent >> amount of history. 3-4GB should be good. I use the t2.medium >> on AWS which has 4GB. >> >> > On May 24, 2017, at 8:51 AM, Arve Knudsen >> <[email protected] <mailto:[email protected]>> wrote: >> > >> > How much memory does the prometheus-k8s-0 pod require? I've >> spun up a Kubernetes cluster of 4 workers with 1 GB of memory >> on DigitalOcean, and prometheus-k8s-0 won't schedule anywhere >> on account of its memory requirements not being met. >> > >> > Thanks, >> > Arve >> > >> > >> >
signature.asc
Description: OpenPGP digital signature
