Use fqdn instead of instance on Redis dashboard panels
The "Redis CPU per Node" dashboard panels try to join metrics from two separate prometheus exporters using the instance
label (which is unique to each prometheus exporter). Consequently those panels are blank.
This MR fixes that by changing the join and aggregation to use the fqdn
label instead of instance
.
Issue gitlab-com/gl-infra/scalability#2240 (closed) documents the problem in more detail.
Snapshot of the updated dashboard's fixed panel:
Edited by Matt Smiley