Network Debug and Troubleshooting Guide¶

Enabling Networking logging and how to read the logfile¶

Platforms: Any

Ansible 2.3 features improved logging to help diagnose and troubleshoot issues regarding Ansible Networking modules.

Because logging is very verbose it is disabled by default. It can be enabled via the ANSIBLE_LOG_PATH and ANSIBLE_DEBUG options on the ansible-controller, that is the machine running ansible-playbook.

Before running ansible-playbook run the following commands to enable logging:

# Specify the location for the log file
export ANSIBLE_LOG_PATH=~/ansible.log
# Enable Debug
export ANSIBLE_DEBUG=True

# Run with 4*v for connection level verbosity
ansible-playbook -vvvv ...

After Ansible has finished running you can inspect the log file which has been created on the ansible-controller:

less $ANSIBLE_LOG_PATH

2017-03-30 13:19:52,740 p=28990 u=fred |  creating new control socket for host veos01:22 as user admin
2017-03-30 13:19:52,741 p=28990 u=fred |  control socket path is /home/fred/.ansible/pc/ca5960d27a
2017-03-30 13:19:52,741 p=28990 u=fred |  current working directory is /home/fred/ansible/test/integration
2017-03-30 13:19:52,741 p=28990 u=fred |  using connection plugin network_cli
...
2017-03-30 13:20:14,771 paramiko.transport userauth is OK
2017-03-30 13:20:15,283 paramiko.transport Authentication (keyboard-interactive) successful!
2017-03-30 13:20:15,302 p=28990 u=fred |  ssh connection done, setting terminal
2017-03-30 13:20:15,321 p=28990 u=fred |  ssh connection has completed successfully
2017-03-30 13:20:15,322 p=28990 u=fred |  connection established to veos01 in 0:00:22.580626

From the log notice:

• p=28990 Is the PID (Process ID) of the ansible-connection process
• u=fred Is the user running ansible, not the remote-user you are attempting to connect as
• creating new control socket for host veos01:22 as user admin host:port as user
• control socket path is location on disk where the persistent connection socket is created
• using connection plugin network_cli Informs you that persistent connection is being used
• connection established to veos01 in 0:00:22.580626 Time taken to obtain a shell on the remote device

Because the log files are verbose, you can use grep to look for specific information. For example, once you have identified the pid from the creating new control socket for host line you can search for other connection log entries:

grep "p=28990" $ANSIBLE_LOG_PATH

Isolating an error¶

Platforms: Any

As with any effort to troubleshoot it’s important to simplify the test case as much as possible.

For Ansible this can be done by ensuring you are only running against one remote device:

• Using ansible-playbook --limit switch1.example.net...
• Using an ad-hoc ansible command

ad-hoc refers to running Ansible to perform some quick command using /usr/bin/ansible, rather than the orchestration language, which is /usr/bin/ansible-playbook. In this case we can ensure connectivity by attempting to execute a single command on the remote device:

ansible -m eos_command -a 'commands=?' -i inventory switch1.example.net -e 'ansible_connection=local' -u admin -k

In the above example, we:

• connect to switch1.example.net specified in the inventory file inventory
• use the module eos_command
• run the command ?
• connect using the username admin
• inform ansible to prompt for the ssh password by specifying -k

If you have SSH keys configured correctly, you don’t need to specify the -k parameter

If the connection still fails you can combine it with the enable_network_logging parameter. For example:

# Specify the location for the log file
export ANSIBLE_LOG_PATH=~/ansible.log
# Enable Debug
export ANSIBLE_DEBUG=True
# Run with 4*v for connection level verbosity
ansible -m eos_command -a 'commands=?' -i inventory switch1.example.net -e 'ansible_connection=local' -u admin -k

Then review the log file and find the relevant error message in the rest of this document.

Error: “[Errno -2] Name or service not known”¶

Platforms: Any

Indicates that the remote host you are trying to connect to can not be reached

For example:

2017-04-04 11:39:48,147 p=15299 u=fred |  control socket path is /home/fred/.ansible/pc/ca5960d27a
2017-04-04 11:39:48,147 p=15299 u=fred |  current working directory is /home/fred/git/ansible-inc/stable-2.3/test/integration
2017-04-04 11:39:48,147 p=15299 u=fred |  using connection plugin network_cli
2017-04-04 11:39:48,340 p=15299 u=fred |  connecting to host veos01 returned an error
2017-04-04 11:39:48,340 p=15299 u=fred |  [Errno -2] Name or service not known

Suggestions to resolve:

• If you are using the provider: options ensure that it’s suboption host: is set correctly.
• If you are not using provider: nor top-level arguments ensure your inventory file is correct.

Error: “Authentication failed”¶

Platforms: Any

Occurs if the credentials (username, passwords, or ssh keys) passed to ansible-connection (via ansible or ansible-playbook) can not be used to connect to the remote device.

For example:

<ios01> ESTABLISH CONNECTION FOR USER: cisco on PORT 22 TO ios01
<ios01> Authentication failed.

Suggestions to resolve:

If you are specifying credentials via password: (either directly or via provider:) or the environment variable ANSIBLE_NET_PASSWORD it is possible that paramiko (the Python SSH library that Ansible uses) is using ssh keys, and therefore the credentials you are specifying are being ignored. To find out if this is the case, disable “look for keys”. This can be done like this:

export ANSIBLE_PARAMIKO_LOOK_FOR_KEYS=False

To make this a permanent change, add the following to your ansible.cfg file:

[paramiko_connection]
look_for_keys = False

Error: “connecting to host <hostname> returned an error” or “Bad address”¶

This may occur if the SSH fingerprint hasn’t been added to Paramiko’s (the Python SSH library) know hosts file.

When using persistent connections with Paramiko, the connection runs in a background process. If the host doesn’t already have a valid SSH key, by default Ansible will prompt to add the host key. This will cause connections running in background processes to fail.

For example:

2017-04-04 12:06:03,486 p=17981 u=fred |  using connection plugin network_cli
2017-04-04 12:06:04,680 p=17981 u=fred |  connecting to host veos01 returned an error
2017-04-04 12:06:04,682 p=17981 u=fred |  (14, 'Bad address')
2017-04-04 12:06:33,519 p=17981 u=fred |  number of connection attempts exceeded, unable to connect to control socket
2017-04-04 12:06:33,520 p=17981 u=fred |  persistent_connect_interval=1, persistent_connect_retries=30

Suggestions to resolve:

Use ssh-keyscan to pre-populate the known_hosts. You need to ensure the keys are correct.

ssh-keyscan veos01

or

You can tell Ansible to automatically accept the keys

Environment variable method:

export ANSIBLE_PARAMIKO_HOST_KEY_AUTO_ADD=True
ansible-playbook ...

ansible.cfg method:

ansible.cfg

[paramiko_connection]
host_key_auto_add = True

Error: “No authentication methods available”¶

For example:

2017-04-04 12:19:05,670 p=18591 u=fred |  creating new control socket for host veos01:None as user admin
2017-04-04 12:19:05,670 p=18591 u=fred |  control socket path is /home/fred/.ansible/pc/ca5960d27a
2017-04-04 12:19:05,670 p=18591 u=fred |  current working directory is /home/fred/git/ansible-inc/ansible-workspace-2/test/integration
2017-04-04 12:19:05,670 p=18591 u=fred |  using connection plugin network_cli
2017-04-04 12:19:06,606 p=18591 u=fred |  connecting to host veos01 returned an error
2017-04-04 12:19:06,606 p=18591 u=fred |  No authentication methods available
2017-04-04 12:19:35,708 p=18591 u=fred |  connect retry timeout expired, unable to connect to control socket
2017-04-04 12:19:35,709 p=18591 u=fred |  persistent_connect_retry_timeout is 15 secs

Suggestions to resolve:

No password or SSH key supplied

Clearing Out Persistent Connections¶

Platforms: Any

In Ansible 2.3, persistent connection sockets are stored in ~/.ansible/pc for all network devices. When an Ansible playbook runs, the persistent socket connection is displayed when verbose output is specified.

<switch> socket_path: /home/fred/.ansible/pc/f64ddfa760

To clear out a persistent connection before it times out (the default timeout is 30 seconds of inactivity), simple delete the socket file.

Timeouts¶

Persistent connection idle timeout:

For example:

2017-04-04 12:19:05,670 p=18591 u=fred |  persistent connection idle timeout triggered, timeout value is 30 secs

Suggestions to resolve:

Increase value of persistent connection idle timeout:

export ANSIBLE_PERSISTENT_CONNECT_TIMEOUT=60

To make this a permanent change, add the following to your ansible.cfg file:

[persistent_connection]
connect_timeout = 60

Command timeout: For example:

2017-04-04 12:19:05,670 p=18591 u=fred |  command timeout triggered, timeout value is 10 secs

Suggestions to resolve:

Options 1: Increase value of command timeout in configuration file or by setting environment variable. Note: This value should be less than persistent connection idle timeout ie. connect_timeout

export ANSIBLE_PERSISTENT_COMMAND_TIMEOUT=30

To make this a permanent change, add the following to your ansible.cfg file:

[persistent_connection]
command_timeout = 30

Option 2: Increase command timeout per task basis. All network modules support a timeout value that can be set on a per task basis. The timeout value controls the amount of time in seconds before the task will fail if the command has not returned.

For example:

Suggestions to resolve:

- name: save running-config
  ios_command:
    commands: copy running-config startup-config
    provider: "{{ cli }}"
    timeout: 30

Some operations take longer than the default 10 seconds to complete. One good example is saving the current running config on IOS devices to startup config. In this case, changing the timeout value form the default 10 seconds to 30 seconds will prevent the task from failing before the command completes successfully. Note: This value should be less than persistent connection idle timeout ie. connect_timeout

Persistent socket connect timeout: For example:

2017-04-04 12:19:35,708 p=18591 u=fred |  connect retry timeout expired, unable to connect to control socket
2017-04-04 12:19:35,709 p=18591 u=fred |  persistent_connect_retry_timeout is 15 secs

Suggestions to resolve:

Increase the value of the persistent connection idle timeout. Note: This value should be greater than the SSH timeout value (the timeout value under the defaults section in the configuration file) and less than the value of the persistent connection idle timeout (connect_timeout).

export ANSIBLE_PERSISTENT_CONNECT_RETRY_TIMEOUT=30

To make this a permanent change, add the following to your ansible.cfg file:

[persistent_connection]
connect_retry_timeout = 30

Error: “invalid connection specified, expected connection=local, got ssh”¶

Platforms: Any

Network modules require that the connection is set to local. Any other connection setting will cause the playbook to fail. Ansible will now detect this condition and return an error message:

fatal: [nxos01]: FAILED! => {
    "changed": false,
    "failed": true,
    "msg": "invalid connection specified, expected connection=local, got ssh"
}

To fix this issue, set the connection value to local using one of the following methods:

• Set the play to use connection: local
• Set the task to use connection: local
• Run ansible-playbook using the -c local setting

Error: “Unable to enter configuration mode”¶

Platforms: eos and ios

This occurs when you attempt to run a task that requires privileged mode in a user mode shell.

For example:

TASK [ios_system : configure name_servers] *****************************************************************************
task path:
fatal: [ios-csr1000v]: FAILED! => {
    "changed": false,
    "failed": true,
   "msg": "unable to enter configuration mode",
}

Suggestions to resolve:

Add authorize: yes to the task. For example:

- name: configure hostname
  ios_system:
    provider:
      hostname: foo
      authorize: yes
  register: result

If the user requires a password to go into privileged mode, this can be specified with auth_pass; if auth_pass isn’t set, the environment variable ANSIBLE_NET_AUTHORIZE will be used instead.

Add authorize: yes to the task. For example:

- name: configure hostname
  ios_system:
  provider:
    hostname: foo
    authorize: yes
    auth_pass: "{{ mypasswordvar }}"
register: result

delegate_to vs ProxyCommand¶

The new connection framework for Network Modules in Ansible 2.3 that uses cli transport no longer supports the use of the delegate_to directive. In order to use a bastion or intermediate jump host to connect to network devices over cli transport, network modules now support the use of ProxyCommand.

To use ProxyCommand, configure the proxy settings in the Ansible inventory file to specify the proxy host.

[nxos]
nxos01
nxos02

[nxos:vars]
ansible_ssh_common_args='-o ProxyCommand="ssh -W %h:%p -q bastion01"'

With the configuration above, simply build and run the playbook as normal with no additional changes necessary. The network module will now connect to the network device by first connecting to the host specified in ansible_ssh_common_args, which is bastion01 in the above example.