TypeError: Cannot read property 'node' of undefined

[jamesbuckett]

Hello.
.
Using this config.js

`var Etcd = require('node-etcd');

var etcdHost = process.env.ETCD_HOST || '172.17.42.1';
var etcdPort = process.env.ETCD_PORT || 4001;
console.log('Connecting to etcd on '+etcdHost+':'+etcdPort);
var etcd = new Etcd(etcdHost, etcdPort);

// Get all Rethinkdb servers from etcd
// We just need one for our connection
var rethinks = etcd.getSync("/services/rethinkdb", {recursive: true})

// Warning: Should add error checking / retry for production use

var nodes = rethinks.body.node.nodes;
nodes.forEach(function (node){
  console.log('Available rethinkdb server on', node.value);
});
var rethink = nodes[0].value;
console.log('Connecting to Rethinkdb on', rethink)

module.exports = {
  rethinkdb: {
    host: rethink,
    port: 28015,
    authKey: '',
    db: 'rethinkdb_ex'
  },
  express: {
     port: 3000
  }
};
`

I run these commands

• fleetctl submit todo@.service todo-sk@.service
• fleetctl list-units
• fleetctl start todo@1 todo-sk@1
• fleetctl journal -f --lines 50 todo@1

I get the following error in the todo@1 log file

Mar 24 11:19:34 core-02 systemd[1]: Started ToDo Service.
Mar 24 11:19:35 core-02 docker[9170]: Connecting to etcd on 172.17.42.1:4001
Mar 24 11:21:41 core-02 docker[9170]: /src/config.js:14
Mar 24 11:21:41 core-02 docker[9170]: var nodes = rethinks.body.node.nodes;
Mar 24 11:21:41 core-02 docker[9170]: ^
Mar 24 11:21:41 core-02 docker[9170]: TypeError: Cannot read property 'node' of undefined
Mar 24 11:21:41 core-02 docker[9170]: at Object.<anonymous> (/src/config.js:14:26)
Mar 24 11:21:41 core-02 docker[9170]: at Module._compile (module.js:426:26)
Mar 24 11:21:41 core-02 docker[9170]: at Object.Module._extensions..js (module.js:444:10)
Mar 24 11:21:41 core-02 docker[9170]: at Module.load (module.js:351:32)
Mar 24 11:21:41 core-02 docker[9170]: at Function.Module._load (module.js:306:12)
Mar 24 11:21:41 core-02 docker[9170]: at Module.require (module.js:361:17)
Mar 24 11:21:41 core-02 docker[9170]: at require (module.js:380:17)
Mar 24 11:21:41 core-02 docker[9170]: at Object.<anonymous> (/src/app.js:6:14)
Mar 24 11:21:41 core-02 docker[9170]: at Module._compile (module.js:426:26)
Mar 24 11:21:41 core-02 docker[9170]: at Object.Module._extensions..js (module.js:444:10)
Mar 24 11:21:41 core-02 docker[9170]: at Module.load (module.js:351:32)
Mar 24 11:21:41 core-02 docker[9170]: at Function.Module._load (module.js:306:12)
Mar 24 11:21:41 core-02 docker[9170]: at Function.Module.runMain (module.js:467:10)
Mar 24 11:21:41 core-02 docker[9170]: at startup (node.js:117:18)
Mar 24 11:21:41 core-02 docker[9170]: at node.js:946:3
Mar 24 11:21:41 core-02 systemd[1]: todo@1.service: Main process exited, code=exited, status=1/FAILURE
Mar 24 11:21:41 core-02 docker[9636]: Error response from daemon: Cannot kill container todo-1: Container df90a15910eb219a081f845839c711a48ad1067c4c11f9707849b9febcb84de5 is not running
Mar 24 11:21:41 core-02 docker[9636]: Error: failed to kill containers: [todo-1]
Mar 24 11:21:41 core-02 docker[9651]: todo-1
Mar 24 11:21:41 core-02 systemd[1]: Stopped ToDo Service.
Mar 24 11:21:41 core-02 systemd[1]: todo@1.service: Unit entered failed state.
Mar 24 11:21:41 core-02 systemd[1]: todo@1.service: Failed with result 'exit-code'.

Thanks In Advance

[rosskukulinski]

Hi @jamesbuckett. Sorry for the confusion! The ToDo service uses rethinkdb as its datastore and there are two other services that you need to run first:
rethinkdb@.service and rethinkdb-announce@.service.

You should be able to submit & start one instance of rethinkdb using these commands:

fleetctl submit rethinkdb@.service rethinkdb-announce@.service
fleetctl start rethinkdb@1.service rethinkdb-announce@1.service

The rethinkdb@.service template runs an instance of the rethinkdb database, while rethinkdb-announce publishes the availability of that rethinkdb instance to etcd (which the ToDo app queries here)

I hope that helps! Let me know if you run into other problems.

[jameszh]

Ross,
I was having the same issue while rethinkdb and rethinkdb-announce were started and up, looked at inside todo@ container, no ETCD_HOST set, so added line in '-e ETCD_HOST=${COREOS_PUBLIC_IPV4}' to "docker run" in the todo@.service, it worked. I'm running "CoreOS stable (899.13.0)".

Thanks,
James

[rosskukulinski]

hi @jameszh. That's interesting. Normally I wouldn't expect ETCD_HOST to be set as an environmental variable -- I used that for local testing in a docker-compose environment.

Here I set var etcdHost = process.env.ETCD_HOST || '172.17.42.1';. Normally CoreOS exposes etcd on the docker network on 172.17.42.1, but maybe that isn't happening in your cluster?

[jameszh]

looks like is the case, my cluster (coreos-vagrant) doesn't have 172.17.42.1, no response from running following command. Must be related to new version of CoreOS.

core@core-01 ~ $ etcdctl -C '172.17.42.1:4001' ls

[rosskukulinski]

ahhhhh! yes that can happen in the default vagrant-coreos configuration because vagrant uses something like the 172.17.x.x subnet internally. I faintly remember mentioning that in the video tutorial series, but it would be good to have an errata / doc for that here

[rosskukulinski]

See: https://github.com/coreos/coreos-vagrant/issues/111 and coreos/coreos-vagrant#209

[rosskukulinski]

also this: https://github.com/coreos/docs/blob/master/etcd/getting-started-with-etcd.md#reading-and-writing-from-inside-a-container

[jamesbuckett]

Thank You Gentlemen

I had to apply both recommendations

1. Edit the Vagrantfile and alter ip = "172.17.8.#{i+100}" to ip = "10.17.8.#{i+100}"
2. vagrant destroy and vagrant up
3. vi todo@.service and add -e ETCD_HOST=${COREOS_PUBLIC_IPV4}' to the RUN section

It is now working.

Thank you for your assistance.