Bug 1378604

Summary: GlusterFS endpoints disappear without a Service in the project
Product: OpenShift Container Platform Reporter: Eric Jones <erjones>
Component: StorageAssignee: Bradley Childs <bchilds>
Status: CLOSED CURRENTRELEASE QA Contact: Jianwei Hou <jhou>
Severity: low Docs Contact:
Priority: medium    
Version: 3.2.1CC: agoldste, aos-bugs, eparis, jsafrane
Target Milestone: ---Keywords: Unconfirmed
Target Release: ---   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2017-02-01 16:04:46 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:

Description Eric Jones 2016-09-22 21:21:13 UTC
Description of problem:
After the upgrade from 3.2.1.4 to 3.2.1.13 for a Red Hat customer, any project with a glusterfs endpoint that did not have a corresponding glusterfs service, the endpoint would disappear.  This link describes the issue well: https://github.com/openshift/origin/pull/6167

Customer indicates that they did not see this issue in 3.2.1.4


Additional info:
Frequently if the service doesn't exist. Creating a service in each project persists the endpoint, which is not ideal since end-users could accidentally delete the service.

Comment 1 Jan Safranek 2016-10-07 14:19:11 UTC
I noticed this OpenShift commit between  3.2.1.4 and 3.2.1.13: https://github.com/openshift/ose/commit/1fc581b329820ea74cde33eac9094cfe91fc332e

Andy, could it lead to missing endpoints in this case?

Comment 2 Eric Paris 2016-10-07 14:23:14 UTC
Isn't this 'working as designed' and one needs to make a headless service to stop the garbage collector?

Comment 3 Andy Goldstein 2016-10-07 14:25:00 UTC
I don't see how. It only modifies endpoints under the "kubernetes" service in the "default" namespace.

Comment 4 Jan Safranek 2016-10-07 14:25:40 UTC
(In reply to Eric Paris from comment #2)
> Isn't this 'working as designed' and one needs to make a headless service to
> stop the garbage collector?

"Customer indicates that they did not see this issue in 3.2.1.4"

I'd expect they will have hard time in 3.3 and 3.4, however we should not break it in a minor release.

Comment 5 Jan Safranek 2017-02-01 16:04:46 UTC
We were unable to reproduce this behavior in 3.2 and we believe that anything related to garbage collection of Endpoints has been fixed in 3.3.