Control Plane latency recommendations for reliable clusters

modules/control-plane-latency.adoc

@@ -0,0 +1,20 @@
+// Module included in the following assemblies:
+//
+// * scalability_and_performance/control-plane-latency-recommendations-for-reliable-clusters.adoc
+
+:_mod-docs-content-type: CONCEPT
+[id="control-plane-latency_{context}"]
+= Recommended Control Plane latency for reliable clusters
+
+Latency between each of the control plane nodes must be less than 15ms to ensure a well performing and reliable cluster. Some of the metrics to keep track of include etcd gRPC requests latency, fysnc latency and any critical alerts. Here are the PromQL queries:
+
++
+[source,terminal]
+----
+histogram_quantile(0.99, sum(rate(grpc_server_handling_seconds_bucket{job=~".*etcd.*", grpc_method!="Defragment", grpc_type="unary"}[5m])) without(grpc_type)) < 0.15
+
+avg_over_time(histogram_quantile(0.99, rate(etcd_disk_wal_fsync_duration_seconds_bucket[2m]))[10m:]) < 1
+
+ALERTS{severity="critical", alertstate="firing"} > 0
+
+----

scalability_and_performance/control-plane-latency-recommendations-for-reliable-clusters.adoc

@@ -0,0 +1,11 @@
+:_mod-docs-content-type: ASSEMBLY
+[id="control-plane-latency-recommendations-for-reliable-clusters"]
+= Control Plane latency recommendations for reliable clusters
+include::_attributes/common-attributes.adoc[]
+:context: control-plane-latency-recommendations-for-reliable-clusters
+
+toc::[]
+
+This topic provides recommended Control Plane latency for reliable clusters.
+
+include::modules/control-plane-latency.adoc[leveloffset=+1]