+Kendall Hartung <[email protected]> +Lee Verberne <[email protected]>
Kendall just started looking into this approach, so your links will be very helpful. Thanks. On Thu, Dec 8, 2016 at 11:16 AM Fabian Reinartz <[email protected]> wrote: > Hey Steve, > > > A second approach might be Prometheus, and running a meta-Prometheus > server to aggregate the individual cluster's results. Has anyone looked > into such an approach before? > > That's exactly the way how you would do it with Prometheus. You would have > a Prometheus instance per cluster that collects metrics and generates > aggregations via recording rules [0]. You can then use federation [1] to > collect those aggregates from the individual per-cluster Prometheus servers > into a global one. > > For running Prometheus on Kubernetes, we also built the Prometheus > Operator [2]. We are keen on any feedback on how we can improve it in > general and for federated setups in particular. > > Fabian > > [0] https://prometheus.io/docs/practices/rules/ > [1] https://prometheus.io/docs/operating/federation/ > [2] https://github.com/coreos/prometheus-operator > > > On Fri, Dec 2, 2016 at 7:23 AM Piotr Szczesniak <[email protected]> > wrote: > > +Steve > > On Wed, Nov 30, 2016 at 5:40 PM, Piotr Szczesniak <[email protected]> > wrote: > > Hi Steve, > > Your use case is out of the scope of Heapster project, which is about > gathering system metrics (like resource usage) rather than a part of > alerting pipeline. Added a few folks from Stackdriver which can elaborate > about potential options with alerts there. Fabian (in cc) should be able to > answer your question regarding Prometheus. > > Thanks, > Piotr > > On Wed, Nov 30, 2016 at 4:30 PM, Vishnu Kannan <[email protected]> wrote: > > +Filip +Piotr > > On Wed, Nov 30, 2016 at 5:31 AM, 'Steve Wolter' via Kubernetes user > discussion and Q&A <[email protected]> wrote: > > Hi kubernetes-users, > > I have the following problem: I need a dashboard and alerts for the health > of my Kubernetes Deployment, for example the # of healthy/unhealthy pods in > the replica sets of the deployment or the observed and actual generation of > the deployment. I'm running half a dozen clusters and my team wants to run > a lot more clusters in the near future, so I'd strongly prefer > cross-cluster aggregation. > > I understand that Heapster is already running in most GKE clusters and > publishing metrics to Stackdriver, so this looks like an ideal point of > integration. Has anyone tried this with Heapster before? Is it a good idea? > Do you know who would be the right persons to talk to for getting health > metrics off the ground in heapster? > > A second approach might be Prometheus, and running a meta-Prometheus > server to aggregate the individual cluster's results. Has anyone looked > into such an approach before? > > Thanks, Steve > > -- > You received this message because you are subscribed to the Google Groups > "Kubernetes user discussion and Q&A" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to [email protected]. > To post to this group, send email to [email protected]. > Visit this group at https://groups.google.com/group/kubernetes-users. > For more options, visit https://groups.google.com/d/optout. > > > > > -- You received this message because you are subscribed to the Google Groups "Kubernetes user discussion and Q&A" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at https://groups.google.com/group/kubernetes-users. For more options, visit https://groups.google.com/d/optout.
