Include Node Exporter in Charts

For Omnibus deployments, we include Node Exporter to be able to retrieve OS and system level metrics. We should consider whether we should include this in the Charts as well.

added Monitoring label

cc @bjk & @stanhu in order to echo comments from discussion during Contribute NOLA 2019.

Personally, I don't think we should for the following reasons:

• Node Exporter in Kubernetes has to run as a daemonset, which means that we would trigger a pod to run on every single node of the cluster, regardless of whether a GitLab pod was running there. For large cluster, this could be duplicative and problematic.
• Node Exporter generally also needs to run privileged, so it can gather additional metrics from the OS/hardware. This can be a security concern.
• We can get metrics like CPU and RAM consumption in other ways in Kubernetes, primarily things like cAdvisor which we do today, which do not rely on deploying a heavy weight solution like a privileged daemonset.

Ultimately, for Kubernetes deployments Node Exporter performs more of cluster monitoring function, rather than an application monitoring function.

For this reason, my gut feel is that we shouldn't include it.

Makes sense to me.

@twk3 I'm strongly tempted to make this closed / wont-fix.

closed

added groupdistribution label and removed 1 deleted label