Technologies/Envoy/envoy.cluster.upstream.connection.total
EnvoyEnvoyMetric

envoy.cluster.upstream.connection.total

Total upstream connections
Dimensions:None
Available on:PrometheusPrometheus (1)OpenTelemetryOpenTelemetry (1)Native (1)
Interface Metrics (3)
PrometheusPrometheus
Total number of upstream connections established to cluster members
Dimensions:None
OpenTelemetryOpenTelemetry
Total number of connections established to upstream cluster members
Dimensions:None
Native
Total number of connections opened to upstream cluster members
Dimensions:None
Sources
envoy_cluster_upstream_cx_totalwww.envoyproxy.io
envoy.cluster.upstream.connection.totalgithub.com
cluster.upstream_cx_totalwww.envoyproxy.io

Technical Annotations (10)

Configuration Parameters (3)
trafficPolicy.connectionPool.http.maxRequestsPerConnectionrecommended: 0
Never close connection based on request count in DestinationRule
trafficPolicy.connectionPool.http.h2UpgradePolicyrecommended: UPGRADE
Multiplex requests over HTTP/2 to reduce handshakes in DestinationRule
trafficPolicy.connectionPool.tcp.tcpKeepalive.timerecommended: 7200s
TCP keepalive time in DestinationRule
CLI Commands (2)
kubectl exec deploy/my-service -c istio-proxy -- pilot-agent request GET stats | grep "downstream_cx_active\|upstream_cx_active"diagnostic
kubectl exec deploy/my-service -c istio-proxy -- pilot-agent request GET stats | grep "downstream_cx_total\|upstream_cx_total"diagnostic
Technical References (5)
upstream clustercomponentvirtual nodecomponentmTLSprotocolDestinationRulecomponentprotocol sniffingconcept
Related Insights (4)
High active connection count indicates connection leakswarning
▸
Upstream connection failures prevent service-to-service communicationwarning
▸
mTLS handshakes are the biggest CPU consumerwarning
▸
Protocol sniffing adds CPU overhead per new connectioninfo
▸