Ansible failing to copy module to target host

Archives Ansible Project
1

Hi,

I’ve ran to a problem provisioning one host.

I have playbook that successfully provisions vagrant box, but running that on amazon host I get this

PUT /tmp/tmpxFGcNu TO /tmp/ansible/ansible-tmp-1395837087.32-71225947296118/setup

ec2-xx-xx-xx-xx.compute-1.amazonaws.com | FAILED => failed to transfer file to /tmp/ansible/ansible-tmp-1395837087.32-71225947296118/setup:

Couldn’t read packet: Connection reset by peer

When I check out tmp dir on remote host I see directiories are created by ansible but no file inside.

Here is complete output:

2

Let me add few more thins and answer to my question.

I have run this on three machines.

Ubuntu 13.04 - Ansible 1.5.3 error was:

PUT /tmp/tmpxFGcNu TO /tmp/ansible/ansible-tmp-1395837087.32-71225947296118/setup

ec2-xx-xx-xx-xx.compute-1.amazonaws.com | FAILED => failed to transfer file to /tmp/ansible/ansible-tmp-

1395837087.32-71225947296118/setup:

Couldn’t read packet: Connection reset by peer

Mac OSX - Ansible 1.4.4:

> FAILED => failed to transfer file to /home/admin/.ansible/tmp/ansible-1395838813.75-19289519445016/setup:

ControlPath "/Users/marko/.ansible/cp/ansible-ssh-ec2-54-236-227-155.compute-1.amazonaws.com-22-admin.MEUoiLFZkgoecYzU" too long for Unix domain socket
subsystem request failed on channel 0
Connection closed

And Debian 7.1 - Ansible 1.6:

failed to open a SFTP connection (Channel closed.)

I have enable scp_if_ssh = True
and all went well.

Is it only version issue woth dofferent error messages?

3

We have been experiencing this bug consistently for some time. Is there a ticket to track this, Ansiblers?

fatal: [ec2-54-xxx-yyy-zz.us-west-2.compute.amazonaws.com] => failed to transfer file to /home/ubuntu/.ansible/tmp/ansible-tmp-1402514262.96-23740295147290/source:

Warning: Permanently added ‘ec2-54-xxx-yyy-zz.us-west-2.compute.amazonaws.com,54.xxx.yyy.zz’ (RSA) to the list of known hosts.
ControlPath “/Users/zaven/.ansible/cp/ansible-ssh-ec2-54-xxx-yyy-xx.us-west-2.compute.amazonaws.com-22-ubuntu.pIvKZuLKKh2LEqwd” too long for Unix domain socket
Write failed: Broken pipe
Connection closed

4

It’s not a bug so much as a setting you need to change in ansible.cfg because there’s very little we can do about it.

See: http://docs.ansible.com/intro_configuration.html#control-path

5

It could be the sftp config line on the host’s /etc/ssh/sshd_config:

Subsystem sftp /usr/libexec/openssh/sftp-server #the path may be different depends on your OS

6

This problem can exist even in cases where the control-path is reconfigured to be shorter. I made that change in my configuration and it did nothing to resolve the issue.

`
<##.#.#.#> ESTABLISH CONNECTION FOR USER: [username]
<##.#.#.#> REMOTE_MODULE setup
<##.#.#.#> EXEC [‘ssh’, ‘-C’, ‘-tt’, ‘-vvv’, ‘-o’, ‘ControlMaster=auto’, ‘-o’, ‘ControlPersist=60s’, ‘-o’, ‘ControlPath=/home/[username]/.ansible/cp/ansible-ssh-%h-%p-%r’, ‘-o’, ‘KbdInteractiveAuthentication=no’, ‘-o’, ‘PreferredAuthentications=gssapi-with-mic,gssapi-keyex,hostbased,publickey’, ‘-o’, ‘PasswordAuthentication=no’, ‘-o’, ‘ConnectTimeout=10’, ‘##.#.#.#’, “/bin/sh -c ‘mkdir -p $HOME/.ansible/tmp/ansible-tmp-1421357644.16-197958593810564 && chmod a+rx $HOME/.ansible/tmp/ansible-tmp-1421357644.16-197958593810564 && echo $HOME/.ansible/tmp/ansible-tmp-1421357644.16-197958593810564’”]
<##.#.#.#> PUT /tmp/tmpXLEUtz TO /home/[username]/.ansible/tmp/ansible-tmp-1421357644.16-197958593810564/setup

fatal: [##.#.#.#] => failed to transfer file to /home/USER/.ansible/tmp/ansible-tmp-1421357538.57-18056896264178/setup:

Received message too long 1919246692

`

As you can see, the number of characters is actually nowhere near the amount needed to trigger that error on my systems, and yet, that’s the error that is reported back.

I think something else is actually going on behind the scenes and the error reported back by ansible is making an assumption about the cause of the failure.

In addition, my /etc/ssh/sshd_config is correct for my OS and the path to sftp-server is correct.

7

Pls ignore. My issue is unrelated and seems to be caused by an “unclean” ssh connection where a .bashrc resource or something similar is echoing something back during the login and causing the content length to be corrupted.

8

The only thing that worked for me was from something I found on Stackoverflow:

My version of it here:

sudo pip install ansible //dont need this if you have it installed of course - I didnt as was using my git clone.
cd /etc/paths.d/
sudo touch ControlPath
sudo echo export ANSIBLE_SSH_CONTROL_PATH=‘%(directory)s/%%h-%%r’ >> ControlPath
source ControlPath