Hi Luis, In my understanding, the bug (4614) is about that the openflowjava cannot handle concurrent barrier requests to the same switch. And the patch (Gerrit 33906) has fixed the bug.
Therefore, I think one consequence of not applying this patch (33906) is that we would need to force applications of OpenFlow plugin not to send barrier requests to the same switch concurrently. The applications need to send a barrier request one by one. I think it's not happy for developers who wants to develop high performance OpenFlow applications. That being said, I think it's good if we can keep to support CPqD. Considering the root cause of this issue is a bug of the CPqD, if CPqD is needed, we should create a patch which not apply the change of the patch (Gerrit 33906) only when openflowjava or someone knows the target switch is CPqD rather than not applying the patch (33906) and force applications to implement complex things. Regards, Hideyuki Tai From: [email protected] [mailto:[email protected]] On Behalf Of Luis Gomez Sent: Friday, February 05, 2016 15:48 To: Michal Polkoráb <[email protected]>; [email protected]; [email protected] Cc: openflowjava-dev <[email protected]>; openflowplugin-dev <[email protected]>; Abhijit Kumbhare <[email protected]> Subject: Re: [vtn-dev] [openflowjava-dev] Regression detected But also, can anyone in vtn explain the consequences of not applying this patch? the reason is this seems like the ovs 2.0 discussion last release: we do something right in the code that breaks popular switch connectivity, and CPqD is not as popular as OVS but still. I am also cc-ing memo project because they have recently showed interest in using CPqD for their system test, and this will impact them. BR/Luis On Feb 5, 2016, at 3:38 PM, Luis Gomez <[email protected]<mailto:[email protected]>> wrote: I can take care of the robot test. On Feb 5, 2016, at 6:48 AM, Michal Polkoráb <[email protected]<mailto:[email protected]>> wrote: Robot tests (robot setups) should know which switch they are using. CPqD switches should be used only when testing groups - as far as I know. Unfortunately, I won't do the change as it is out of my scope. Michal ________________________________ From: Abhijit Kumbhare <[email protected]<mailto:[email protected]>> Sent: 05 February 2016 15:36 To: Michal Polkoráb Cc: Luis Gomez; openflowjava-dev; openflowplugin-dev Subject: Re: [openflowjava-dev] Regression detected For the first solution: 1. introduce timeout / sleep in testcases with CPqD switches, How do you know if we are connecting to a CPqD switch and not any other kind of switch? It will be important to be able to keep working with CPqD switch as it would be good to be able to work with multiple switches. So will you do the timeout change Michal? On Fri, Feb 5, 2016 at 5:43 AM, Michal Polkoráb <[email protected]<mailto:[email protected]>> wrote: Hello, the change that causes the problems is this one: https://git.opendaylight.org/gerrit/#/c/33906/. It fixes Bug 4614 (https://bugs.opendaylight.org/show_bug.cgi?id=4614). We have tested the change with both OVS and CPqD switches and both passed. The XID (see the change) wasn't used because CPqD switch didn't fully behave according to the specification (didn't reply with correct XID) - and as it turns out, it still doesn't. The first connection usually replies with random generated XID in features-reply, after we close the connection (approx. 30 seconds), the device connects successfully. So our options are: 1. introduce timeout / sleep in testcases with CPqD switches, 2. replace CPqD switches with OVS switches (since OVS 2.3 - groups are supported). Regards, Michal Polkorab ________________________________ From: Abhijit Kumbhare <[email protected]<mailto:[email protected]>> Sent: 05 February 2016 01:24 To: Luis Gomez Cc: openflowjava-dev; openflowplugin-dev Subject: Re: [openflowjava-dev] Regression detected Thanks Luis. May be better to wait for Michal Polkorab to comment. On Thu, Feb 4, 2016 at 4:08 PM, Luis Gomez <[email protected]<mailto:[email protected]>> wrote: Hi OFP devs, It looks like this patch: https://git.opendaylight.org/gerrit/#/c/33652/ broke CPqD switch support in both plugins: https://jenkins.opendaylight.org/releng/view/openflowplugin/job/openflowplugin-csit-1node-flow-services-only-beryllium/ https://jenkins.opendaylight.org/releng/view/openflowplugin/job/openflowplugin-csit-1node-flow-services-lithium-redesign-only-beryllium/ BR/Luis _______________________________________________ openflowjava-dev mailing list [email protected]<mailto:[email protected]> https://lists.opendaylight.org/mailman/listinfo/openflowjava-dev MichalPolkoráb Software Engineer Sídlo / Mlynské Nivy 56 / 821 05 Bratislava / Slovakia R&D centrum / Janka Kráľa 9 / 974 01 Banská Bystrica / Slovakia +421 918 378 907<tel:%2B421%20918%20378%20907> / [email protected]<mailto:[email protected]> reception: +421 2 206 65 114 / www.pantheon.sk<http://www.pantheon.sk/> [Image removed by sender. logo] MichalPolkoráb Software Engineer Sídlo / Mlynské Nivy 56 / 821 05 Bratislava / Slovakia R&D centrum / Janka Kráľa 9 / 974 01 Banská Bystrica / Slovakia +421 918 378 907 / [email protected]<mailto:[email protected]> reception: +421 2 206 65 114 / www.pantheon.sk<http://www.pantheon.sk/> [Image removed by sender. logo]
_______________________________________________ openflowplugin-dev mailing list [email protected] https://lists.opendaylight.org/mailman/listinfo/openflowplugin-dev
