tripleo

hosts-config.sh reports Argument list too long

Bug #1856076 reported by Luke Short on 2019-12-11

This bug affects 2 people

Affects		Status	Importance	Assigned to	Milestone
	tripleo	Fix Released	High	Luke Short	tripleo ussuri-1

Bug Description

Description
===========
When deploying a large Overcloud (over 400 nodes in my case), the deployment fails during step0 due to this error: "OSError: [Errno 7] Argument list too long"

It happens during this task:

```
$ less deploy_steps_playbook.yaml
    - name: HostsConfig
      script: hosts-config.sh
      environment:
        HOSTS_ENTRY: "{{ hosts_entry }}"

```

Steps to reproduce
==================
1. Deploy a large Overcloud

Expected result
===============
The script should complete successfully.

Actual result
=============
The script fails causing the entire deployment to fail.

Environment
===========
Train release
Pre-deployed servers

Logs & Configs
==============
http://paste.openstack.org/show/787468/

Luke Short (ekultails) on 2019-12-11

Changed in tripleo:
importance:	Undecided → High

Revision history for this message

Luke Short (ekultails) wrote on 2019-12-11:

The underlying problem is that the HOSTS_ENTRY environment variable is passed to the remote host and it is huge because it contains a list of every Overcloud host. It follows in the format of:

Here is a truncated example, as reported by the debug output of Ansible: http://paste.openstack.org/show/787479/

Revision history for this message

Oliver Walsh (owalsh) wrote on 2019-12-17:

We dad a similar issue with ssh_known_hosts in the past (which is even bigger) https://bugs.launchpad.net/tripleo/+bug/1772071

Revision history for this message

Luke Short (ekultails) wrote on 2019-12-23:

Fixed via: https://review.opendev.org/#/c/669965/