fleetctl ssh fails with Vagrant

[coreos]

fleetctl ssh fails with Vagrant

[bcwaldon]

...aaaaand I screwed up. The docs for the environment variable stuff landed before the actual feature :) Use --tunnel for now.

[joelmoss]

Well everything else works. I can list-units, list-machines, start, stop, etc. with any problem.

[joelmoss]

And passing --tunnel also fails, but with a slightly different error...

$ fleetctl --tunnel 127.0.0.1:2222 ssh --unit hello.service
2014/02/20 20:52:43 Unable to establish SSH connection: ssh: unable to open direct tcpip connection: open failed

[bcwaldon]

Is Vagrant binding etcd to your laptop's localhost port 4001?

[bcwaldon]

And are you confident Vagrant is using port 2222 for that SSH tunnel? Maybe share the output of

vagrant ssh-config

[joelmoss]

Is Vagrant binding etcd to your laptop's localhost port 4001?

yup

[joelmoss]

$ vagrant ssh-config
Host default
  HostName 127.0.0.1
  User core
  Port 2222
  UserKnownHostsFile /dev/null
  StrictHostKeyChecking no
  PasswordAuthentication no
  IdentityFile /Users/joelmoss/.vagrant.d/insecure_private_key
  IdentitiesOnly yes
  LogLevel FATAL

list-units works...

$ fleetctl list-units                                      
UNIT        LOAD    ACTIVE  SUB DESC    MACHINE
hello.service   loaded  active  running -   fb3d1c42...

[bcwaldon]

For the fleetctl list-units (and other working commands), fleetctl is just talking to etcd bound to localhost port 4001.

For the commands that do not work (status, journal, etc), fleetctl wants to SSH to the machine hosting a given unit. Your machine is not publishing an IP address, so it is probably trying to ssh to localhost port 22, which is refusing connections.

[bcwaldon]

I'm going to go poke at this on my box real quick.

[bcwaldon]

@joelmoss What's the output of fleetctl list-machines?

[joelmoss]

$ fleetctl list-machines                   
MACHINE     IP  METADATA
fb3d1c42... -   -

[bcwaldon]

One more... cat /etc/os-release

[bcwaldon]

So the reason your machine isn't publishing an IP is because fleet just looks for the first ipv4 address on the eth0 interface, which vagrant does not have.

[joelmoss]

core@localhost ~ $ cat /etc/os-release
NAME=CoreOS
ID=coreos
VERSION=231.0.0
VERSION_ID=231.0.0
BUILD_ID=
PRETTY_NAME="CoreOS 231.0.0 (Official Build) dev-channel amd64-generic test"
ANSI_COLOR="1;32"
HOME_URL="http://www.coreos.com/"

[bcwaldon]

Thank you, sir. I'll get back to you in a minute.

[joelmoss]

Using my Vagrant private IP also fails...

$ fleetctl --tunnel 172.12.8.150 ssh --unit hello.service 
2014/02/20 21:03:55 Unable to establish SSH connection: ssh: unable to open direct tcpip connection: open failed

[bcwaldon]

@joelmoss #152 will enable us to fix this problem on vagrant

[bcwaldon]

We can use that code with fleet.conf:

public_interface="lo"

...but we really need the stock coreos vagrant box to ship with it

[bcwaldon]

I actually have a better solution in #155. It uses netlink to determine which local interface is likely to have a routable IP bound to it. When using it with Vagrant, you end up with the IP on the private vagrant network, but that's ok:

% fleetctl --tunnel 127.0.0.1:2222 list-units
UNIT        LOAD    ACTIVE  SUB DESC    MACHINE
hello.service   loaded  active  running -   d66f8269.../10.0.2.15
% fleetctl --tunnel 127.0.0.1:2222 status hello.service
hello.service - Hello World
   Loaded: loaded (/run/systemd/system/hello.service; enabled-runtime)
   Active: active (running) since Thu 2014-02-20 14:06:43 UTC; 1min 18s ago
 Main PID: 4092 (bash)
   CGroup: /system.slice/hello.service
           ├─4092 /bin/bash -c while true; do echo "Hello, world"; sleep 1; done
           └─4198 sleep 1

Feb 20 14:07:52 localhost bash[4092]: Hello, world
Feb 20 14:07:53 localhost bash[4092]: Hello, world
Feb 20 14:07:54 localhost bash[4092]: Hello, world
Feb 20 14:07:55 localhost bash[4092]: Hello, world
Feb 20 14:07:56 localhost bash[4092]: Hello, world
Feb 20 14:07:57 localhost bash[4092]: Hello, world
Feb 20 14:07:58 localhost bash[4092]: Hello, world
Feb 20 14:07:59 localhost bash[4092]: Hello, world
Feb 20 14:08:00 localhost bash[4092]: Hello, world
Feb 20 14:08:01 localhost bash[4092]: Hello, world