Recovery from site.pp misconfig for control node

Bug #1166740 reported by Paul Michali
10
This bug affects 2 people
Affects Status Importance Assigned to Milestone
Cisco Openstack
Triaged
Undecided
Unassigned
Grizzly
Triaged
Low
Unassigned

Bug Description

In setting up site.pp, the node name for the control node was changed everywhere (e.g. control-101), except for $controller_hostname, which had the default value of control-server. After running "puppet apply" and "clean_node.sh control-101", the site.pp file was corrected so that $controller_hostname matched the other occurrences of the host name.

Running "cobbler system list" showed the correct node name (control-101), and "clean_node.sh control-101" proceeded to build the node. However, the node was created with the name control-server and since there is no site.pp node definition for that node, it uses a default node type and doesn't set up the node as a control node.

Attempted to resolve by "cobbler system remove --name=control-101", "puppet cert clean control-server", dropping and recreating the puppet database in MySQL, and re-running "puppet apply" and "clean_node.sh control-101", but the same result occurred.

Found that issue was caused by /etc/hosts containing an entry for control-server with the same IP, and positioned before the entry for control-101. Apparently this file is being used to obtain the node name, instead of using what is in site.pp.

Although a misconfiguration, this was both difficult to recover from and not obvious as to why the control-server was still being used, even when all traces of it were removed from the scripts, databases, etc.

Changed in openstack-cisco:
milestone: 2012.2.4 → none
Changed in openstack-cisco:
status: New → Triaged
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.