Hi Arun,

Take a look at [6] In case of OVSDB we only use one version and one model, most 
recent, so we used this fix.

You may want to look at Tom’s patch [7] as OFP uses network-topology and 
inventory both and [6] may not work. Once Robert’s patch lands you can revert 
yours.

Regards,
Vishal.

[6] https://git.opendaylight.org/gerrit/67191
[7] 
https://git.opendaylight.org/gerrit/#/c/66536/2/impl/src/main/java/org/opendaylight/jsonrpc/impl/Util.java

From: D Arunprakash
Sent: 18 January 2018 08:46
To: Luis Gomez <[email protected]>; Vishal Thapar <[email protected]>; 
openflowplugin-dev <[email protected]>
Subject: RE: [openflowplugin-dev] [opendaylight-dev] [release] odlparent + 
yangtools bump

Hi Luis,
We haven’t started anything on this, let me check with Vishal and comeback on 
this.

Regards,
Arun

From: 
[email protected]<mailto:[email protected]>
 [mailto:[email protected]] On Behalf Of Luis 
Gomez
Sent: Thursday, January 18, 2018 12:45 AM
To: Vishal Thapar 
<[email protected]<mailto:[email protected]>>; 
openflowplugin-dev 
<[email protected]<mailto:[email protected]>>
Subject: Re: [openflowplugin-dev] [opendaylight-dev] [release] odlparent + 
yangtools bump

Is anyone looking at this? this is currently blocking netvirt and other 
projects relying in ofp.

BR/Luis

On Jan 16, 2018, at 7:43 PM, Vishal Thapar 
<[email protected]<mailto:[email protected]>> wrote:

Hi Luis,

There is already a Jira [4], and patch [5] from Robert in works. OFPlugin could 
do a similar workaround to one I did for OVSDB while waiting for Robert’s patch 
to land.

Regards,
Vishal.

[4] https://jira.opendaylight.org/browse/NETCONF-496
[5] https://git.opendaylight.org/gerrit/67212

From: 
[email protected]<mailto:[email protected]>
 [mailto:[email protected]] On Behalf Of Luis 
Gomez
Sent: 17 January 2018 08:58
To: openflowplugin-dev 
<[email protected]<mailto:[email protected]>>
Subject: [openflowplugin-dev] Fwd: [opendaylight-dev] [release] odlparent + 
yangtools bump

FYI the topology issue described by Vishal below also happens in 
openflowplugin, so the proposed workarounds should also work here. Should we 
open a ticket for this?

BR/Luis


Begin forwarded message:

From: Vishal Thapar 
<[email protected]<mailto:[email protected]>>
Subject: Re: [opendaylight-dev] [release] odlparent + yangtools bump
Date: January 16, 2018 at 6:08:46 AM PST
To: Tom Pantelis <[email protected]<mailto:[email protected]>>
Cc: Stephen Kitt <[email protected]<mailto:[email protected]>>, odl-dev-list 
<[email protected]<mailto:[email protected]>>, 
"[email protected]<mailto:[email protected]>"
 
<[email protected]<mailto:[email protected]>>,
 Release <[email protected]<mailto:[email protected]>>

Hi Tom,

Yeah, this is another workaround. In case of OVSDB I know which version I want 
to use. But it is still not a proper fix. Something like restconf/apidocs can’t 
know the correct version. If latest is correct one, should probably move older 
ones to differently named bundles.

Regards,
Vishal.

From: Tom Pantelis [mailto:[email protected]]
Sent: 16 January 2018 19:13
To: Vishal Thapar 
<[email protected]<mailto:[email protected]>>
Cc: Robert Varga <[email protected]<mailto:[email protected]>>; Stephen Kitt 
<[email protected]<mailto:[email protected]>>; odl-dev-list 
<[email protected]<mailto:[email protected]>>; Release 
<[email protected]<mailto:[email protected]>>; 
[email protected]<mailto:[email protected]>
Subject: Re: [release] [opendaylight-dev] odlparent + yangtools bump



On Tue, Jan 16, 2018 at 1:57 AM, Vishal Thapar 
<[email protected]<mailto:[email protected]>> wrote:
Thanks for the lead Tom.

I’m not subscribed to yangtools-dev so they may not get this mail.

I followed it to new yangtools unable to figure out which version to pick when 
2 versions of same model are loaded. Network-topology happens to be one of 8 
such instances I saw that have two different versions under same yang file and 
both are loaded [1]. I think this is more of an existing bad design or 
oversight being exposed by new yangtools, your opinion may wary.

I was able to workaround part of OVSDB issue by hardcoding the revision when 
deserializing [2]. I am now able to use OVSDB plugin to configure OVS, but 
restconf is still failing as well as odl-mdsal-apidocsc gives 500 error. IMO, 
there are two ways to take care of this:


  1.  Publish and modify all test scripts etc. to specify revision in rest 
calls too [if possible].
  2.  Eliminate duplicate models by moving older models to appropriately named 
folders, something similar to ietf-yang-types [3] only moving older one to use 
the revision in folder name, to avoid changes to all consumers.
  3.  ‘Fix’ in yangtools to pick the newer version.

Note that I mention yangtools but this could likely be an mdsal or restconf 
issue too, am not familiar enough with any of these to make a definitive 
statement on where the ‘bug’ is. Also, without rest allOVSDB/Genius/Netvirt 
CSIT will be broken, even if rest of bump activity goes through smoothly.


This sounds familiar - I ran into it in another project and worked around it: 
https://git.opendaylight.org/gerrit/#/c/66536/2/impl/src/main/java/org/opendaylight/jsonrpc/impl/Util.java
 - the comments in the findModuleWithLatestRevision method sum it up.


Thanks and Regards,
Vishal.

[1] 
https://github.com/opendaylight/mdsal/tree/master/model/ietf/ietf-topology/src/main/yang
[2] https://git.opendaylight.org/gerrit/67191
[3] https://github.com/opendaylight/mdsal/tree/master/model/ietf


From: Tom Pantelis [mailto:[email protected]<mailto:[email protected]>]
Sent: 15 January 2018 19:34
To: Vishal Thapar 
<[email protected]<mailto:[email protected]>>
Cc: Stephen Kitt <[email protected]<mailto:[email protected]>>; odl-dev-list 
<[email protected]<mailto:[email protected]>>; Release 
<[email protected]<mailto:[email protected]>>
Subject: Re: [release] [opendaylight-dev] odlparent + yangtools bump

org.opendaylight.ovsdb.southbound.InstanceIdentifierCodec.prefixForNamespace 
returned null:

    @Override
    protected String prefixForNamespace(final URI namespace) {
        return context.findModule(namespace, 
Optional.empty()).map(Module::getName).orElse(null);
    }




On Mon, Jan 15, 2018 at 8:27 AM, Vishal Thapar 
<[email protected]<mailto:[email protected]>> wrote:
Hi all,

I know bump is still in progress but I've seen basic functionality issues in 
OVSDB which means all projects depending on OVSDB are also broken. Main issue 
I've seen:

1. The issue is at startup, OVSDB plugin creates default topology for ovsdb:1 
and hwvtep:1 in network-topology config and operational. They both don't show 
up when I make a restconf GET call. It was working earlier.

2. I have genius code to write to network-topology config. It triggers Data 
change notification and I can see entry in logs showing so, but it throws up 
this error:

2018-01-15T17:56:42,977 | ERROR | 
opendaylight-cluster-data-notification-dispatcher-176 | 
DataTreeChangeListenerActor      | 137 - 
org.opendaylight.controller.sal-clustering-commons - 1.7.0.SNAPSHOT | 
member-1-shard-topology-config: Error notifying 
listenerorg.opendaylight.ovsdb.southbound.OvsdbDataTreeChangeListener@3eeb42f<mailto:org.opendaylight.ovsdb.southbound.OvsdbDataTreeChangeListener@3eeb42f>
 java.lang.IllegalArgumentException: Failed to map QName {} 
[(urn:TBD:params:xml:ns:yang:network-topology?revision=2013-10-21)network-topology]
  at com.google.common.base.Preconditions.checkArgument(Preconditions.java:210) 
[62:com.google.guava:23.3.0.jre]
  at 
org.opendaylight.yangtools.yang.data.util.AbstractNamespaceCodec.appendQName(AbstractNamespaceCodec.java:47)
 [294:org.opendaylight.yangtools.yang-data-util:2.0.0]
  at 
org.opendaylight.yangtools.yang.data.util.AbstractStringInstanceIdentifierCodec.serialize(AbstractStringInstanceIdentifierCodec.java:53)
 [294:org.opendaylight.yangtools.yang-data-util:2.0.0]
  at 
org.opendaylight.ovsdb.southbound.InstanceIdentifierCodec.serialize(InstanceIdentifierCodec.java:65)
 [270:org.opendaylight.ovsdb.southbound-impl:1.6.0.SNAPSHOT]
  at 
org.opendaylight.ovsdb.southbound.ovsdb.transact.TransactUtils.stampInstanceIdentifierMutation(TransactUtils.java:440)
 [270:org.opendaylight.ovsdb.southbound-impl:1.6.0.SNAPSHOT]
  at 
org.opendaylight.ovsdb.southbound.ovsdb.transact.TerminationPointCreateCommand.stampInstanceIdentifier(TerminationPointCreateCommand.java:372)

So looks like model is loaded correctly, data is also written/read correctly, 
but something going wrong with the Codec or base model making it fail some 
checks.


-----Original Message-----
From: 
[email protected]<mailto:[email protected]> 
[mailto:[email protected]<mailto:[email protected]>]
 On Behalf Of Stephen Kitt
Sent: 12 January 2018 22:03
To: odl-dev-list 
<[email protected]<mailto:[email protected]>>; Release 
<[email protected]<mailto:[email protected]>>
Subject: Re: [opendaylight-dev] [release] odlparent + yangtools bump

On Fri, 12 Jan 2018 11:05:01 +0100
Michael Vorburger <[email protected]<mailto:[email protected]>> wrote:
> FYI for anyone else looking at the published meeting minutes from
> yesterday's TSC call which decided doing this so late in the game for
> Oxygen:
>
> https://meetings.opendaylight.org/opendaylight-meeting/2018/tsc/openda
> ylight-meeting-tsc.2018-01-11-18<tel:2018-01-11-18>.00.html
> https://meetings.opendaylight.org/opendaylight-meeting/2018/tsc/openda
> ylight-meeting-tsc.2018-01-11-18<tel:2018-01-11-18>.00.txt
> https://meetings.opendaylight.org/opendaylight-meeting/2018/tsc/openda
> ylight-meeting-tsc.2018-01-11-18<tel:2018-01-11-18>.00.log.html
>
> Question: was there a succesful CSIT run of the netvirt project, as
> discussed as a pre-requisite during the kernel project calls on
> Tuesday?

No, we ran into various issues with the jobs on Jenkins and the patches which 
meant I couldn’t get a CSIT run before the TSC call. We decided, instead of 
spending more time trying to get a CSIT run *before* the merge, to go ahead 
with a time-boxed merge and run CSIT then — the time-box is that we want CSIT 
on Monday at the latest.

It would no doubt have taken me longer than that to get CSIT working on the 
basis of the multi-patch jobs, so apart from the disruption caused to projects 
today, this seems like the less intrusive approach...
(Well, still more intrusive than just dropping the patches, but the TSC wants 
the migration to happen if at all possible anyway.)

Regards,

Stephen
_______________________________________________
release mailing list
[email protected]<mailto:[email protected]>
https://lists.opendaylight.org/mailman/listinfo/release


_______________________________________________
dev mailing list
[email protected]<mailto:[email protected]>
https://lists.opendaylight.org/mailman/listinfo/dev

_______________________________________________
openflowplugin-dev mailing list
[email protected]
https://lists.opendaylight.org/mailman/listinfo/openflowplugin-dev

Reply via email to