Not able to plot the cluster level metrics, if run on multiple nodes

[Dileep-Dora]

We have a multi nodes cluster running and have installed elasticsearch_exporter on all of them, but while running the below dashboard getting error for cluster-level panels.

https://grafana.com/grafana/dashboards/2322

Only queries that return a single series/table is supported - this is the error we're getting on Grafana.

Upon checking query resulting two values for that, so to monitor all nodes, using --es.all is the way or we can install on individual nodes and run as well, if yes in dashboard how we can get the cluster level panels.

[alexon1234]

Same here

[snoby]

@Dileep-Dora - did you come to a resolution on this?

[caarlos0]

you don't need to run it on all nodes, the exporter will still get metrics from all nodes anyway...

ideally I think you should run one exporter per cluster.

[ahmedasimabdelhamid]

did you managed to make it work?

[Tristan971]

@caarlos0

you don't need to run it on all nodes, the exporter will still get metrics from all nodes anyway...

ideally I think you should run one exporter per cluster.

Well yes and no... There is an explicit opt-in with --es.all that is supposedly defaulting to false... If the exporter forcefully monitors all nodes even when unset, that might be something one wants to specify.

Additionally, running an ES cluster (perhaps partially for redundancy) but monitoring from a single node (if you don't run some proxy in front of it since many application clients support passing multiple node hosts for example) is at best a bit odd and will make prometheus metrics suggest the whole cluster is dead during a rolling restart when it's not the case, for example...