the Query interface

Basically what this example does, is first retrieving all metrics that match the filters, and showing them on stacked graphs, one graph for the metrics for device bond0, and another graph for bond1. The metrics are speeds in b/s, averaged out over 4 matching servers. The graph shows a plot for each type (rx and tx).

How this works in detail, I'll explain below:

• filters Only retrieve targets that match the filters. Searches in the metric id, or on the given tags, either with substring matching, regex search or equality checks. Negation is supported and order doesn't matter. All targets_matching metrics that match the filters, will be used, upto limit.

• group_by
  control how metrics are grouped into graphs: graphs bundle all targets that have the same value for all tags in group_by. If metrics have a different value for any of these tags, they go to a different graph. By default grouping happens by target_type, unit, and server; you can extend or replace this list; in this case we group by target_type, unit and device. So all graphs_from_matching_targets have a unique selection of values of these tags. See also constant tags.

• avg_by
  one of the aggregation methods: for all targets that go together on a graph, all targets with any value for this tag will be averaged and represented as one. I.e. the graph shows the average of all 4 matched servers. You can also aggregate by summing.

• avg_over
  depending on how spikey a graph is, you can smooth it out by averaging over a certain time interval.

• from_to
  control the time range.

• constant tags
  these tags are the same for all metrics of this graph; either explicitly because of group_by, or by "coincidence" (e.g. in this example they happen to have the same value for "plugin" (network))

• variable tags
  targets on the graph are represented by those tags that distinguish themselves from the others on this graph. In this case the only difference between the metrics on the graph is type=rx vs type=tx.

• tag_legend
  shows all tag keys found amongst the found metrics. the color of the tag is generated from the tag key, so because usually only the value of tags is shown (i.e. in constant tags and variable tags), this helps in recognizing the key of tags, though you can alsways mouseover any tag to see the key

• vtitle
  shows the unit tag.

• other_graphs
  because some targets have another value for one of the group_by tags is (device is bond1), GE shows those on a new graph.