Bug 1376578

Summary: remove kubernetes from 7.3; retain etcd + flannel
Product: Red Hat Enterprise Linux 7 Reporter: Micah Abbott <miabbott>
Component: rhel-server-atomicAssignee: Colin Walters <walters>
Status: CLOSED CURRENTRELEASE QA Contact: atomic-bugs <atomic-bugs>
Severity: unspecified Docs Contact:
Priority: unspecified    
Version: 7.3CC: bbreard, eparis, jlebon, lfriedma, miabbott, rcyriac, smilner
Target Milestone: rcKeywords: Extras
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: 2017-05-31 09:37:24 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 Micah Abbott 2016-09-15 19:09:04 UTC
When composing the 7.3 trees, we should be removing the 'kubernetes' package, but keeping 'etcd' and 'flannel'

I really have no justification for this, other than word of mouth.

If I can find a more concrete requirement, I'll share it.

Comment 4 Colin Walters 2016-09-16 12:05:14 UTC
https://lists.projectatomic.io/projectatomic-archives/atomic/2015-November/msg00003.html

As far as flannel...basically, the theme here is we stop shipping multi-machine/clustering software by default, leaving that to higher layers.  flannel definitely falls in that category, as does gluster/ceph.

Comment 5 Colin Walters 2016-09-16 12:06:05 UTC
Let's not forget here that part of this whole idea is that now etcd is just an `atomic host install` away.

Comment 8 Colin Walters 2016-09-21 13:08:21 UTC
So we're just ignoring the comment from Subhendu I linked in comment #4 ?

Why aren't we editing the PRD or whatever is driving the filing of this bug?

Comment 9 Eric Paris 2016-09-21 13:34:26 UTC
That's correct Colin, I am 'ignoring' that comment. If some document needs changed I hope Ben and/or Subhendu will change it. etcd may need to start before flannel which needs to start before docker. Which means system container. Which last I heard was being called tech preview. Even if we call it fully supported removing etcd was shown to break OpenShift. So that's a non-starter.

We will continue to work towards the utopian goal, but we aren't there.

Comment 10 Micah Abbott 2016-09-21 13:38:10 UTC
This was discussed in an email thread with lfriedma, bbreard, eparis, imcleod, and others.

The consensus was that the PRD only calls for the kube apiserver, controller-mgr, and scheduler to be removed.  It was also agreed that we were not prepared to deal with the removal of etcd or flannel.

When asked about the status of this bug with regards to the PRD, imcleod said:

"You can safely assume that the PRD and the requirements on RHEL AH for 7.3.0 will not change.  The bug reflects the fact that the composition of the host changed in a way that was not requested or expected for 7.3.0, which is an entirely reasonable thing to flag."

Comment 11 Colin Walters 2016-09-21 14:01:23 UTC
(In reply to Eric Paris from comment #9)
> That's correct Colin, I am 'ignoring' that comment. If some document needs
> changed I hope Ben and/or Subhendu will change it. etcd may need to start
> before flannel which needs to start before docker. Which means system
> container.

Or layered packages.  Which I did test.

> Which last I heard was being called tech preview. Even if we call
> it fully supported removing etcd was shown to break OpenShift. So that's a
> non-starter.

Reference?  I'm guessing this is something like etcd_container ended up relying on /etc/etcd/ being present, which is pretty easy to fix.

Comment 13 Micah Abbott 2016-09-22 14:01:27 UTC
Latest 7.3 snap4 compose contains these requested packages.

Comment 14 Jon Disnard 2017-05-31 09:37:24 UTC
(In reply to Micah Abbott from comment #13)
> Latest 7.3 snap4 compose contains these requested packages.

Then I'm moving this to closed current