Issue with labelmaps in prometheus relabel config generating empty strings

[eldondev]

I hope you don't mind if I err on the side of conciseness in this issue. If it would be better I can edit this to match the github template. The relabel labelmap configs such as are defined here seem to use the nodename as the regex to match. However, the "replacement" component of the "labelmap" instance takes the first regex capture group if I am reading this correctly. Since no capture group is present, the result is an empty label (""="node1.local" for example). This causes alertmanager to complain to me. However in my case the examples that I found had an "instance" label, that I would expect to match the "node name" label. In my case, I believe removing the additional labelmap step was sufficient to generate desired behavior. What was the labelmap intended to do?
before removing relabel:
etcd_debugging_disk_backend_commit_rebalance_duration_seconds_count{="node1.local",instance="node1.local",job="etcd"}
after:
etcd_debugging_disk_backend_commit_rebalance_duration_seconds_count{instance="node1.local",job="etcd"}

If you'd like, I can create a PR just to remove the misbehaving labelmaps.

[dghubble]

Yeah, that's right. You can remove that labelmap.

In its original form, all node labels were collected. But time series reductions removed that since those labels are seldom valuable (#596) leaving just the name. But the instance is what's used in dashboards.