only add prom autopilot gauges to servers #11241
Conversation
acpana
force-pushed
the
ffmmm/b-10730-client-fix
branch
from
c29092c
to
096fde6
Compare
Signed-off-by: FFMMM <[email protected]>
acpana
force-pushed
the
ffmmm/b-10730-client-fix
branch
from
096fde6
to
202495f
Compare
acpana
added
backport/1.10
pr/do-not-merge
| @@ -216,6 +216,11 @@ func getPrometheusDefs(cfg lib.TelemetryConfig) ([]prometheus.GaugeDefinition, [ | |||
| raftGauges, | |||
| } | |||
|
|
|||
| // TODO(ffmmm): conditionally add only leader specific metrics to gauges, counters, summaries, etc | |||
There was a problem hiding this comment.
+1 to this comment
Seems like it would be worth it to move all the server-only metrics in this PR, since they are all here in this same area.
There was a problem hiding this comment.
Yeah, agreed that we have the opportunity to separate our metrics. And also communicate that on the docs.
I'd rather us wait for hashicorp/go-metrics#129 to go in.
That way, we can programmatically test over all the metrics for the desired behavior. ie appear on servers, not appear on clients.
I can write out a quick issue around this.
For now, I'd love for us to merge this in so we can fully close #10730
|
🍒 If backport labels were added before merging, cherry-picking will start automatically. To retroactively trigger a backport after merging, add backport labels and re-run https://circleci.com/gh/hashicorp/consul/471498. |
|
🍒✅ Cherry pick of commit 62980ff onto |
Signed-off-by: FFMMM <[email protected]>
Overview
This PR makes it such that consul autopilot gauges are not configured for
PrometheusOptsfor non server agents (clients).For clients, there is no documented behavior on https://www.consul.io/docs/agent/telemetry#autopilot. In
1.9consul emittedNaNfor these metrics. In1.10, by virtue of changes to the underlyinggo-metricslib, consul emits0.This change proposes not emitting these metrics at all for clients.
Issue Related
consul.autopilot.healthy|consul_autpilot_healthymetric in Consul 1.10 #10730