Bug 977075 - nova [Negative]: no time out when instances fail to build
Summary: nova [Negative]: no time out when instances fail to build
Keywords:
Status: CLOSED WONTFIX
Alias: None
Product: Red Hat OpenStack
Classification: Red Hat
Component: openstack-nova
Version: unspecified
Hardware: x86_64
OS: Linux
medium
medium
Target Milestone: ---
: 6.0 (Juno)
Assignee: Vladan Popovic
QA Contact: Ami Jeain
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2013-06-23 09:52 UTC by Dafna Ron
Modified: 2019-09-09 13:37 UTC (History)
7 users (show)

Fixed In Version:
Doc Type: Enhancement
Doc Text:
Clone Of:
Environment:
Last Closed: 2014-06-12 20:51:57 UTC
Target Upstream Version:


Attachments (Terms of Use)
logs (3.04 MB, application/x-gzip)
2013-06-23 11:14 UTC, Dafna Ron
no flags Details

Description Dafna Ron 2013-06-23 09:52:54 UTC
Description of problem:

I noticed that when there is a problem with instances they are stuck in build forever with no timeout. 
this has happened on my setup for a cinder issue (after restart of cinder the problem was fixed and might be related to build update) but I also reproduced it on other scenario in which I launched 10 instances and rebooted the host.
some of the instances were stuck in build state forever. 
 
perhaps it will be a good idea to create a time-out for status build and move instances to status Error

Version-Release number of selected component (if applicable):

openstack-nova-compute-2013.1.2-2.el6ost.noarch
openstack-nova-api-2013.1.2-2.el6ost.noarch

How reproducible:

100%

Steps to Reproduce:
1. launch 10 instances
2. reboot the host
3.

Actual results:

some vms move to error, some are stuck in build. 

Expected results:

we should have a time-out and move instances to Error state 

Additional info: logs

Comment 1 Dafna Ron 2013-06-23 11:14:29 UTC
Created attachment 764291 [details]
logs

Comment 2 Nikola Dipanov 2013-11-01 14:39:33 UTC
In most cases - if something is wrong with the instance - it should go to ERROR - however in cases like rebooting the host - it can happen that they stay in BUILD. 

Currently - the sync_power_states periodic task does not consider BUILD instances, and having a timeout (as opposed to just checking state as the periodic task does now) so this will require an upstream blueprint.

Moving to 5.0.

Comment 4 Stephen Gordon 2014-04-06 20:18:04 UTC
Vladan is this something that you have actively looked at or do I need to move to 6.0?

Comment 5 Vladan Popovic 2014-04-17 12:44:51 UTC
Sorry for the late response Stephen. 
Didn't manage to look at this still, so it should be moved to 6.0.

Comment 6 Russell Bryant 2014-06-12 20:51:57 UTC
Please report an upstream bug for this.  The upstream bug will need to contain a lot more detail about how to reproduce this scenario, though.


Note You need to log in before you can comment on or make changes to this bug.