Invalid Prometheus metric names

[pschultz]

With at least one TLS cluster enabled Envoy produces invalid output on the /stats?format=prometheus endpoint, such as this one:

# TYPE envoy_cluster_ssl_ciphers_ECDHE-RSA-AES128-GCM-SHA256 counter
envoy_cluster_ssl_ciphers_ECDHE-RSA-AES128-GCM-SHA256{envoy_cluster_name="local_service"} 8

Dashes are not allowed. The cipher suite should be put in a label instead, e.g.

# TYPE envoy_cluster_ssl_ciphers_ECDHE-RSA-AES128-GCM-SHA256 counter
envoy_cluster_ssl_ciphers_used_total{envoy_cluster_name="local_service", cipher_suite="ECDHE-RSA-AES128-GCM-SHA256"} 8

Repro steps:

• launch Envoy with config below
• send at least one request to the L1 listener: curl -I localhost:8080
• request /stats?format=prometheus on the admin listener: curl -s localhost:5000/stats?format=prometheus | grep ciphers

Config:

admin:
  access_log_path: /dev/null
  address:
    socket_address: { address: 0.0.0.0, port_value: 5000 }

static_resources:
  listeners:
  - name: L1
    address:
      socket_address: { address: 0.0.0.0, port_value: 8080 }
    filter_chains:
     - filters:
       - name: envoy.http_connection_manager
         config:
           stat_prefix: egress_http
           codec_type: AUTO
           http_filters:
           - name: envoy.router
           route_config:
             virtual_hosts:
             - name: google
               domains: ['*']
               routes: [{ match: { prefix: / }, route: { cluster: "google" } }]

  clusters:
  - name: google
    connect_timeout: 1s
    type: STRICT_DNS
    dns_lookup_family: V4_ONLY
    lb_policy: ROUND_ROBIN
    tls_context: {}
    hosts:
    - socket_address: { address: www.google.com, port_value: 443 }

[mattklein123]

Fixed