Bug 826116 - Conductor shouldn't attempt to deploy if configserver is unavailable
Summary: Conductor shouldn't attempt to deploy if configserver is unavailable
Keywords:
Status: CLOSED CURRENTRELEASE
Alias: None
Product: CloudForms Cloud Engine
Classification: Retired
Component: aeolus-conductor
Version: 1.0.0
Hardware: Unspecified
OS: Unspecified
unspecified
high
Target Milestone: 1.0.1
Assignee: Angus Thomas
QA Contact: dgao
URL:
Whiteboard:
Depends On: 809621
Blocks:
TreeView+ depends on / blocked
 
Reported: 2012-05-29 16:11 UTC by Chris Pelland
Modified: 2012-12-13 19:50 UTC (History)
9 users (show)

Fixed In Version: aeolus-conductor-0.8.20-1
Doc Type: Bug Fix
Doc Text:
Clone Of: 809621
Environment:
Last Closed: 2012-12-13 19:50:06 UTC
Embargoed:


Attachments (Terms of Use)
down_cs (242.77 KB, image/png)
2012-06-11 22:29 UTC, dgao
no flags Details

Description Chris Pelland 2012-05-29 16:11:08 UTC
+++ This bug was initially created as a clone of Bug #809621 +++

Description of problem:

If a working configserver becomes inactive, any attempts to launch a new application deployment will result in stuck deployments.  The stuck deployments cannot be stopped or deleted.  To prevent this, conductor should check to see if the configserver is operational, before launching.

Version-Release number of selected component (if applicable):
 * aeolus-conductor-0.8.7-1.el6.src.rpm
 * aeolus-configure-2.5.2-1.el6.src.rpm
 * imagefactory-1.0.0rc11-1.el6.src.rpm
 * oz-0.8.0-5.el6.src.rpm
 * rubygem-aeolus-cli-0.3.1-1.el6.src.rpm
 * rubygem-aeolus-image-0.3.0-12.el6.src.rpm

How reproducible:
 * 2 out of 2 attempts

Steps to Reproduce:
1. Install and configure Aeolus conductor capable
2. Deploy and configure a working configserver
3. Update the cloud provider account information with valid configserver information
4. Make the configserver go away (block all traffic with iptables, or shut it down)
5. Attempt to launch an application that relies on configserver
  
Actual results:

The UI provides the following notifications:

> Warnings
> Failed to launch following component blueprints:

> Errors
> systemNo route to host - connect(2)

 * At this point, conductor shows a deployment in the 'new' state.  It never leaves that state, and I cannot delete the application.

Expected results:

I'd expect to either ...
 1) not be allow to deploy when the cfgserver is out of reach
 2) or, be able to delete failed deployments that resulted from missing cfgserver

Additional info:

 * See attached debug tarball

--- Additional comment from whayutin on 2012-04-03 16:42:50 EDT ---

related to https://bugzilla.redhat.com/show_bug.cgi?id=796528 possibly

--- Additional comment from jprovazn on 2012-05-23 10:41:30 EDT ---

I believe that patch for https://bugzilla.redhat.com/show_bug.cgi?id=796528 fixes this too.

--- Additional comment from matt.wagner on 2012-05-23 15:12:58 EDT ---

Confirmed -- the patch for #796528 does resolve this issue. With an unreachable config server, instances go directly to create_failed state. It's on master, but not backported anywhere yet. I'm setting this to "modified" to match that bug.

--- Additional comment from matt.wagner on 2012-05-25 13:59:55 EDT ---

The relevant commits on https://bugzilla.redhat.com/show_bug.cgi?id=796528 are:
7a8502b846a819c27fa141621220ca0bbaeac23c
56016671e651cf17bb0bc5c29b49c5aa55e94536
3dd5f304b8458528d15ddf462e0db7622b56dd09
86987cd9194c344272c0cfff313edcdf66df80c0

Though it sounds like QE isn't pleased with 796528 yet.

Comment 1 Angus Thomas 2012-05-31 16:34:52 UTC
This bug is believed to have been resolved by the patch that was applied as part of bz #826130

Comment 2 dgao 2012-06-11 22:29:00 UTC
Created attachment 591034 [details]
down_cs

[root@qeblade38 ~]# rpm -qa | grep "aeolus"
aeolus-conductor-doc-0.8.27-1.el6_3.noarch
aeolus-conductor-daemons-0.8.27-1.el6_3.noarch
aeolus-conductor-0.8.27-1.el6_3.noarch
rubygem-aeolus-image-0.3.0-12.el6.noarch
aeolus-configure-2.5.7-1.el6_3.noarch
aeolus-all-0.8.27-1.el6_3.noarch
rubygem-aeolus-cli-0.3.3-1.el6_3.noarch

Conductor return a create_failed state and error message when configserver got shutdown.

Verified.


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