Ajay Tripathy

@cliffcolvin if these are memory improvements we should help keep this branch up-to-date and help add tests. Can we get an internal issue filed?

Any chance there are docs on deploying the dcgm exporter?

@thomasvn could you possibly pick up testing this as part of your GPU sprint?

@thomasvn if this is meaningful I may just ask you to take over the fix?

In that case what do we need to merge @thomasvn ?

@mhrabovcin you can look here for how its done in kubecost: https://github.com/kubecost/cost-analyzer-helm-chart/tree/develop/cost-analyzer/grafana-dashboards As Matt Ray noted, multicluster is supported by hooking up to any promql endpoint that can aggregate data...