srvaddr

Small command-line utility for querying SRV DNS records and rendering the results into a template or an easily manipulated JSON structure.

Installation

The usual:

$ go get -u github.com/pd/srvaddr
$ go install github.com/pd/srvaddr

You can download binaries from the Github Releases:

$ curl -o srvaddr https://github.com/pd/srvaddr/releases/download/v0.1/srvaddr_linux_amd64 && \
    chmod +x ./srvaddr && \
    ./srvaddr -h

Alternatively, it's available as a docker image:

$ docker run philodespotos/srvaddr _xmpp-server._tcp.google.com
xmpp-server.l.google.com:5269
alt1.xmpp-server.l.google.com:5269
alt2.xmpp-server.l.google.com:5269
alt3.xmpp-server.l.google.com:5269
alt4.xmpp-server.l.google.com:5269

Examples

The default template simply lists the hostname(s) and port(s) returned for the given query:

$ srvaddr _frontend._tcp.tld
box2.example.com:1234
box3.example.com:7654
box1.example.com:9876

The results are displayed in no particular order.

NB: Nothing I use in practice actually populates the priority or weight values in SRV records, but the results should probably be sorted based on those.

To lookup multiple services at once, the default unlabeled output is ambiguous. You can label each service and use the "env-style" template:

$ srvaddr ZK=_zookeeper._tcp.service.consul MQ=_rabbit._tcp.service.consul
ZK_ADDR0=spof.node.dc1.consul:2181
ZK_HOST0=spof.node.dc1.consul
ZK_PORT0=2181
ZK_ADDR1=spof.node.dc1.consul:2181
ZK_HOST1=spof.node.dc1.consul
ZK_PORT1=2181
MQ_ADDR0=spof.node.dc1.consul:5672
MQ_HOST0=spof.node.dc1.consul
MQ_PORT0=5672

Under the hood, the output formats are all just defined with text/template. You can use a custom template with -t filename (with - representing stdin, as usual):

$ echo '{{with index .db 0}}database port is {{.Port}}{{end}}' | \
    srvaddr -t - db=_postgres._tcp.internal
database port is 16273

Go's templating can be unwieldy at times. You can just emit JSON, maybe pipe it into jq:

$ srvaddr -json zk=_zookeeper._tcp.service.consul | jq .
{
  "zk": [
    {
      "Host": "deadbeef.node.dc1.consul.",
      "IP": "10.2.2.2",
      "Port": 2181
    },
    {
      "Host": "beadfeed.node.dc1.consul.",
      "IP": "10.1.1.1",
      "Port": 2181
    }
  ]
}

Not all DNS servers will return the additional A records necessary for the IP to be known; favor using the hostname, which is part of the SRV record itself, whenever possible.

By default, srvaddr uses a nameserver from resolv.conf; to use a different nameserver:

$ srvaddr -ns 127.0.0.1:8600 _kafka._tcp.service.consul

Development

I'll probably write tests at some point. Setting up the mocks was too distracting early on. Meanwhile, a fairly low friction way of serving up some SRV records is to run consul locally. Here's a minimal example config:

{
  "server": true,
  "bootstrap_expect": 1,
  "log_level": "debug",
  "services": [
    { "id": "zk1", "name": "zk", "address": "10.1.1.1", "port": 2181 },
    { "id": "zk2", "name": "zk", "address": "10.2.2.2", "port": 2181 },
    { "name": "api", "address": "10.3.3.3", "port": 443 },
    { "name": "mq", "address": "127.0.0.1", "port": 5672 }
  ]
}