Bug 1417150

Summary: Refreshing aws provider and then deleting makes provider items not deleted
Product: Red Hat CloudForms Management Engine Reporter: Matouš Mojžíš <mmojzis>
Component: ProvidersAssignee: Marcel Hild <mhild>
Status: CLOSED DUPLICATE QA Contact: Matouš Mojžíš <mmojzis>
Severity: high Docs Contact:
Priority: high    
Version: 5.7.0CC: dajohnso, gblomqui, jfrey, jhardy, mmojzis, obarenbo
Target Milestone: GA   
Target Release: 5.9.0   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard: provider:ec2:ui
Fixed In Version: Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2017-07-24 17:18:14 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Attachments:
Description Flags
evm.log
none
aws.log none

Description Matouš Mojžíš 2017-01-27 10:06:49 UTC 
Description of problem:


Version-Release number of selected component (if applicable):
5.7.1.0

How reproducible:
Always

Steps to Reproduce:
1. Add an ec2 provider
2. Wait for full refresh and then do refresh again
3. Immediately after that delete provider

Actual results:
Items like VMs/key pairs and other are not deleted from vmdb.

Expected results:
Items should be deleted.

Additional info:
Comment 2 Greg Blomquist 2017-02-02 15:01:43 UTC 
Matouš, do the items show up as archived?  If so, then it's working as designed.

MIQ doesn't delete objects like VMs when a provider is deleted because of the overhead of deleting related records like Metrics and Events.
Comment 3 Matouš Mojžíš 2017-02-02 17:05:01 UTC 
How can I find out whether it's archived?
For example when I do this and add another ec2 provider and try to provision an instance then I have all the images twice in image selection. I think this is not how it should be.
Comment 4 Marcel Hild 2017-03-14 13:28:27 UTC 
I'm pretty sure this is a race condition.
We should schedule the delete of a provider.

https://github.com/ManageIQ/manageiq/pull/13204 evaluates this approach.

Matouš can you do this again and attach all logs to this BZ?

related BZs
https://bugzilla.redhat.com/show_bug.cgi?id=1369359
https://bugzilla.redhat.com/show_bug.cgi?id=1343328
Comment 5 Matouš Mojžíš 2017-03-14 20:26:20 UTC 
Created attachment 1263043 [details]
evm.log
Comment 6 Matouš Mojžíš 2017-03-14 20:27:45 UTC 
Whole evm.log, fresh appliance and I only added, refreshed, deleted ec2 provider. Version 5.7.1.3.
Also aws log attached.
Comment 7 Matouš Mojžíš 2017-03-14 20:28:08 UTC 
Created attachment 1263044 [details]
aws.log
Comment 8 Marcel Hild 2017-07-18 08:48:18 UTC 
reading the logs, this is indeed because the destroy of the provider was running at the same time as a refresh and thus was creating inventory (templates) while deleting it leaving disconnected entities.

since https://github.com/ManageIQ/manageiq/pull/14848 we have an orchestrated destroy, which waits for all workers to stop and then does the destroy.

Matouš I just checked with a current upstream appliance and it worked as expected.
The only "issue" I had was, that I needed to explicitly destroy the dependent Network and storage managers. 
After that, the cloud manager got destroyed and left no templates or vms.

Ok to close this?
Comment 9 Greg Blomquist 2017-07-24 17:18:14 UTC 
Dup'ing this on bug #1437549.  I know this one is older, but that one has more details about the fix and is tracking the relevant PRs.

*** This bug has been marked as a duplicate of bug 1437549 ***