Skip to content

Latest commit

 

History

History
163 lines (109 loc) · 6.65 KB

advanced.md

File metadata and controls

163 lines (109 loc) · 6.65 KB
Raw

Advanced Metrics

There are two Advanced modes (see Metric Modes) which include all Basic Metrics plus extra metrics providing Pod-Level context.

The two Advanced modes are remote context and local context. The difference lies in the Context Labels. Additionally, local context supports Annotations.

Prefix

All metrics have the prefix networkobservability_.

Universal Labels

Node and Cluster metadata are included with the labels:

  • cluster
  • instance (Node name)

Context Labels

There are Pod-Level context labels for metrics prepended with adv_.

To customize context labels, see MetricsConfiguration CRD.

Remote Context

For Advanced mode with remote context, default context labels are the following:

  • source_ip
  • source_namespace
  • source_pod
  • source_workload
  • destination_ip
  • destination_namespace
  • destination_pod
  • destination_workload

Local Context

For Advanced mode with local context, default context labels are the following for outgoing traffic:

  • source_ip
  • source_namespace
  • source_pod
  • source_workload

For incoming traffic:

  • destination_ip
  • destination_namespace
  • destination_pod
  • destination_workload

List of Metrics

Plugin: packetforward (Linux)

Same metrics as Basic mode.

Plugin: dropreason (Linux)

Metrics enabled when dropreason plugin is enabled (see Metrics Configuration).

Metric Name Description Extra Labels
drop_count Basic: dropped packet count direction, reason
drop_bytes Basic: dropped byte count direction, reason
adv_drop_count Advanced/Pod-Level: dropped packet count direction, reason, context labels
adv_drop_bytes Advanced/Pod-Level: dropped byte count direction, reason, context labels

Label Values

See Context Labels.

Possible values for direction:

  • ingress (incoming traffic)
  • egress (outgoing traffic)

Possible values for reason:

  • IPTABLE_RULE_DROP
  • IPTABLE_NAT_DROP
  • TCP_CONNECT_BASIC
  • TCP_ACCEPT_BASIC
  • TCP_CLOSE_BASIC
  • CONNTRACK_ADD_DROP
  • UNKNOWN_DROP

Plugin: linuxutil (Linux)

Same metrics as Basic mode.

Plugin: dns (Linux)

Metrics enabled when dns plugin is enabled (see Metrics Configuration).

Metric Name Description Extra Labels
dns_request_count Basic: number of DNS requests by query query_type, query
dns_response_count Basic: number of DNS responses by query, error code, and response value query_type, query, return_code, response, num_response
adv_dns_request_count Advanced/Pod-Level: number of DNS requests by query query_type, query, context labels
adv_dns_response_count Advanced/Pod-Level: number of DNS responses by query, error code, and response value query_type, query, return_code, response, num_response, context labels

Plugin: hnsstats (Windows)

Same metrics as Basic mode.

Plugin: packetparser (Linux)

Metrics enabled when packetparser plugin is enabled (see Metrics Configuration).

Metric Name Description Extra Labels
adv_forward_count Advanced/Pod-Level: forwarded packet count direction, context labels
adv_forward_bytes Advanced/Pod-Level: forwarded byte count direction, context labels
adv_tcpflags_count Advanced/Pod-Level: TCP packet count by flag flag, context labels
adv_node_apiserver_latency Advanced: API Server round trip time for SYN-ACK (histogram) le (histogram bucket)
adv_node_apiserver_no_response Advanced: number of packets that did not get a response from API server
adv_node_apiserver_tcp_handshake_latency Advanced: API Server latency in establishing connection (histogram) le (histogram bucket)

Note: API Server metrics help identify degradation of Node-to-API-server connection. The metrics were born out of a real-life incident, where Node-to-API-server latency was the root cause.

Label Values

See Context Labels.

Possible values for direction:

  • ingress (incoming traffic)
  • egress (outgoing traffic)

Possible values for flag:

  • FIN
  • SYN
  • RST
  • PSH
  • ACK
  • URG
  • ECE
  • CWR
  • NS

Possible values for le (for API server metrics). Units are in milliseconds. le stands for "less than or equal". See Prometheus histogram documentation for more info.

  • 0
  • 0.5
  • 1 through 4.5 in increments of 0.5
  • inf

Plugin: tcpretrans (Linux)

Metrics enabled when tcpretrans plugin is enabled (see Metrics Configuration).

Metric Name Description Extra Labels
adv_tcpretrans_count Advanced/Pod-Level: TCP retransmitted packet count context labels

Label Values

See Context Labels.