Bug 1326798 (vdsm_config_NetworkMgr_to_be_passive)

Summary: [RFE] Run Vdsm while NM is running
Product: [oVirt] vdsm Reporter: Fabian Deutsch <fdeutsch>
Component: GeneralAssignee: Edward Haas <edwardh>
Status: CLOSED CURRENTRELEASE QA Contact: Meni Yakove <myakove>
Severity: medium Docs Contact:
Priority: high    
Version: 4.17.10CC: bugs, cshao, danken, dguo, fdeutsch, green, huzhao, leiwang, mavital, mburman, myakove, rbarry, sbonazzo, stirabos, ycui, ylavi, yzhao
Target Milestone: ovirt-4.1.0-betaKeywords: FutureFeature
Target Release: 4.19.2Flags: rule-engine: ovirt-4.1+
rule-engine: blocker+
mavital: testing_plan_complete-
ylavi: planning_ack+
danken: devel_ack+
myakove: testing_ack+
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: Enhancement
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2017-03-16 14:46:14 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: Network RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Bug Depends On: 1344411, 1345919, 1346947, 1347958, 1367261, 1389324, 1400874, 1408219, 1420708    
Bug Blocks: 1140646, 1160423, 1304509, 1330138, 1330144, 1337050    

Description Fabian Deutsch 2016-04-13 12:26:34 UTC
Description of problem:
To my understanding NM can be running alongside of vdsm, it should just not mess with devices by itself.

This can be achieved (IIUIC) by setting the no-auto-default=* configuration directive.

If this works, then the benefit would be that NM could run alongside vdsm and projects integrating with NM (like Cockpit) would continue to work.

Comment 1 Fabian Deutsch 2016-04-13 12:31:34 UTC
Note: If this works, then we do not need to disbale NM during host-deploy.

Comment 2 Fabian Deutsch 2016-04-13 12:32:28 UTC
Adjusting priority to high, because having NM enabled is currently blocking the normal he-flow.

he-setuo expects NM to be disabled, and this is conflicting with the idea of using Cockpit for initial network configuration.

Comment 3 Dan Kenigsberg 2016-04-14 11:51:42 UTC
Can you try if pulling NetworkManager-config-server does the trick?

Comment 4 Ryan Barry 2016-04-14 13:10:04 UTC
It does not. hosted-engine setup checks for the explicitly. Simone is testing right now.

Comment 5 Simone Tiraboschi 2016-04-15 11:34:49 UTC
First attempt to deploy hosted-engine with active NetworkManager on the simplest case (just one interface, static addressing) was OK.

Danken, should we completely remove the NetworkManager check on hosted-engine side cause we are sure that VDSM is fully compatible with it?

Comment 6 Fabian Deutsch 2016-04-15 11:58:22 UTC
One thing to note here is that it seems thast vdsm still requires the legacy network.service.

Dan, would it be possible to let NM handle the ifcfg files?

Comment 7 Dan Kenigsberg 2016-04-18 08:43:30 UTC
(In reply to Simone Tiraboschi from comment #5)
> Danken, should we completely remove the NetworkManager check on
> hosted-engine side cause we are sure that VDSM is fully compatible with it?

I'm not sure what does hosted-engine-setup do with regards to NM, but if Vdsm works fine in parallel to NM, we should stop disabling NM in ovirt-host-deploy and elsewhere.

(In reply to Fabian Deutsch from comment #6)
> One thing to note here is that it seems thast vdsm still requires the legacy
> network.service.
> 
> Dan, would it be possible to let NM handle the ifcfg files?

I think that what you're suggesting is writing an NM configurator to Vdsm. This is possible, but not on our roadmap for 4.0.

Comment 8 Fabian Deutsch 2016-04-18 09:25:02 UTC
(In reply to Dan Kenigsberg from comment #7)
> (In reply to Fabian Deutsch from comment #6)
> > One thing to note here is that it seems thast vdsm still requires the legacy
> > network.service.
> > 
> > Dan, would it be possible to let NM handle the ifcfg files?
> 
> I think that what you're suggesting is writing an NM configurator to Vdsm.
> This is possible, but not on our roadmap for 4.0.

No, I'm not suggesting to write an NM configurator.

I'm suggesting to let NM handle the ifcfg files.
Currentl the network.service (legacy) is enforced by vdsm to handle the ifcfg files, but NM can do this as well (with the - by default enabled - ifcfg-rh plugin)

Comment 9 Dan Kenigsberg 2016-04-18 10:30:03 UTC
(In reply to Fabian Deutsch from comment #8)

> I'm suggesting to let NM handle the ifcfg files.
> Currentl the network.service (legacy) is enforced by vdsm to handle the
> ifcfg files, but NM can do this as well (with the - by default enabled -
> ifcfg-rh plugin)

We used to have nasty races with NM_CONTROLLED=yes; we should research if they still exist.

Comment 11 Dan Kenigsberg 2016-04-28 11:59:38 UTC
I'm granting devel_ack for allowing NM to run side-by-side to Vdsm. We are NOT going to manager networking via NM in 4.0.

Comment 12 Sandro Bonazzola 2016-05-02 10:05:22 UTC
Moving from 4.0 alpha to 4.0 beta since 4.0 alpha has been already released and bug is not ON_QA.

Comment 13 Fabian Deutsch 2016-05-06 09:28:05 UTC
*** Bug 1330128 has been marked as a duplicate of this bug. ***

Comment 14 Sandro Bonazzola 2016-05-20 07:48:47 UTC
All referenced patches are merged, is something missing to keep this in POST?

Comment 15 Dan Kenigsberg 2016-05-23 12:32:51 UTC
Meni, could you approve testing that Vdsm is working fine when NM is running?

Comment 16 Meni Yakove 2016-05-29 14:17:32 UTC
Tier1 PASS with NM up and with the patch

Comment 17 Fabian Deutsch 2016-06-03 21:20:24 UTC
Simone, I'm actually missing a patch in master which is changing the default to ohostedcons.NetworkEnv.REFUSE_DEPLOYING_WITH_NM to False.

Do you plan to add this for master and ovirt-4.0, now with the backing in comment 16?

Comment 18 Dan Kenigsberg 2016-06-14 14:08:42 UTC
This bug has failed QA due to bug 1344411: NM is not really passive.

Comment 19 Meni Yakove 2016-08-10 10:27:25 UTC
Network tier2 pass with NM service enabled but this bug failed QE on bug 1344411 which in NEW status. Should we verify this bug or not?

Comment 20 Yaniv Lavi 2016-08-10 11:58:10 UTC
NM is not passive yet. We still have issue with DNS. For now don't verify.

Comment 21 Michael Burman 2016-08-10 12:29:56 UTC
So it shouldn't be ON_QA, please set it to depends on or move to modified.

Comment 22 Red Hat Bugzilla Rules Engine 2016-09-05 09:09:02 UTC
Target release should be placed once a package build is known to fix a issue. Since this bug is not modified, the target version has been reset. Please use target milestone to plan a fix for a oVirt release.

Comment 23 Dan Kenigsberg 2016-09-05 10:29:44 UTC
The best we can have in 4.0.z is consuming NM-controlled interfaces (bug 1304509); I am not sure we can keep NM running.

Comment 24 Dan Kenigsberg 2016-10-26 10:11:33 UTC
Edy is now working on consuming NM-created interfaces while NM is running, slated for 4.0.7.

Comment 25 Simone Tiraboschi 2017-01-05 09:03:58 UTC
*** Bug 1406820 has been marked as a duplicate of this bug. ***

Comment 26 Dan Kenigsberg 2017-01-17 12:37:53 UTC
Nothing to be done on ovirt side for 4.1.

Comment 27 Meni Yakove 2017-03-08 09:28:42 UTC
rhevm-4.1.1.3-0.1.el7.noarch