SECTOR 03 // REFERENCE

tcp

A connect/reachability probe: is the port up, and how long does connecting take?

L4 reachability. Dials a TCP address and reports whether the port accepts a connection and how long that took. It is protocol-agnostic, so it probes any backend by its socket alone: a cache, a queue, a database, a gRPC service. The connect latency in meta is a degradation lens that needs no application client.

providers:
  cache:
    source: tcp
    config:
      addr: "localhost:6379"   # or host + port below
      # host: localhost
      # port: 6379
      # timeout: 5             # connect deadline in seconds (default 5)

The configured target is the default; any step may override it. Give an addr (host:port), or a host plus port pair, and an optional timeout connect deadline in seconds.

Verbs #

connect (probe) #

Dials the address once and times the connect. A reachable port succeeds; a refused or timed-out connection is a probe failure, so connect works as a steadyState gate and inside wait_until.

arg	type	req	description
`addr`	string	no	target `host:port`; defaults to the configured target (primary)
`host`	string	no	target host, paired with `port` instead of `addr`
`port`	number	no	target port, paired with `host`
`timeout`	number	no	connect deadline in seconds (default 5)

Returns true. meta.connectMs (int) is the connect latency and meta.addr (string) is the target dialed. output is "connected to <addr> in <n>ms".

steadyState:
  - run: cache.connect            # the configured target

verify:
  - run: cache.connect
    with: "db:5432"               # a different address
    as: c
  - run: assert
    with: { of: "${.outputs.c.meta.connectMs}", lt: 50 }
    desc: "connects in under 50ms"