Red Hat Bugzilla – Bug 963282
[oo-accept-node] httpd config references DNS name without associated gear
Last modified: 2013-11-17 19:47:30 EST
Description of problem:
When cleaning up after the recent release and changes to accept-node we have been seeing these errors:
FAIL: httpd config references DNS name without associated gear: 'pfpcore-apsp001.rhcloud.com'
At some point in time in the past the namespace has changed and left behind this artifact. The new url is pfpcore-hampton.rhcloud.com.
When namespaces change we need to ensure that we are cleaning up the old urls.
I have seen at least 10-15 of these exact errors.
Version-Release number of selected component (if applicable):
2.0.27 just went out the door.
I'm not sure as most of the creation dates for these are back in march but its definitely worth trying to reproduce.
Steps to Reproduce:
1. Create an application under a certain namespace.
2. Change the namespace of the account.
3. Verify the config was properly removed from the frontend apache. Can do this by running oo-accept-node -v and see if you see the error.
Artifacts left behind.
Remove the old artifacts and swap the new namespace into the frontend apache.
These are extremely painful and require manual cleanup.
It looks as though we are no longer supporting namespace changes to domains which contain applications.
# rhc domain update rmillner0100 0010renllimr
Changing namespace 'rmillner0100' to '0010renllimr' ... Domain contains applications. Delete applications first before changing the domain namespace.
And the code to do namespace updates was removed from the FrontendHttpServer class in the previous release (2.0.27).
The rhc-fix-stale-frontend script should help and can take multiple names on the command line.
What's the contents of .env for the gear in question? I'm concerned this is another instance of an app which has ancient variable names.
This application no longer exists. I see logs on two ex-lg-node(7 & 9) servers but no remains of this application.
If this is not allowed any more than I suggest we move this to closed.
Agreed - upstream can no longer express this bug since the feature has been removed.