Hi guys,
we know that FRS is currently slower than FRM, but a week ago we refactored
with Miroslav the robot test locally to sleep more than 1 or 3 seconds
just because we wanted to eliminate the failures caused by perfromance issues.
We tried to run tests for FRM and FRS with OVS versions 2.0/2.3/2.4/2.5
and we see both are failing on same flows with particular ovs versions -
example from todays testing:
flow: odl/test/csit/variables/xmls/f18.xml
log:set DEBUG org.opendaylight.openflowplugin.impl.services.SalFlowServiceImpl
feature:install odl-openflowplugin-flow-services-rest
OVS 2.0
2016-10-04 10:46:58,499 | DEBUG | entLoopGroup-7-2 | SalFlowServiceImpl
| 209 - org.opendaylight.openflowplugin.impl - 0.4.0.SNAPSHOT | Flow add
with id=141 finished without error
OVS 2.5
2016-10-04 10:50:17,148 | DEBUG | entLoopGroup-7-4 | SalFlowServiceImpl
| 209 - org.opendaylight.openflowplugin.impl - 0.4.0.SNAPSHOT | Flow add
failed for flow=AddFlowInput [_cookie=FlowCookie [_value=18],
_cookieMask=FlowCookie [_value=255], _flowName=FooXf18, _flowRef=FlowRef
[_value=KeyedInstanceIdentifier{targetType=interface
...
_strict=false, augmentation=[]], errors=Device reported error type BADMATCH
code BADWILDCARDS
feature:install odl-restconf; feature:install
odl-openflowplugin-app-config-pusher; feature:install
odl-openflowplugin-app-topology; feature:install
odl-openflowplugin-app-forwardingrules-sync;
OVS 2.0
2016-10-04 10:55:02,697 | DEBUG | entLoopGroup-9-5 | SalFlowServiceImpl
| 281 - org.opendaylight.openflowplugin.impl - 0.4.0.SNAPSHOT | Flow add
with id=141 finished without error
OVS 2.5
2016-10-04 10:54:08,438 | DEBUG | entLoopGroup-9-3 | SalFlowServiceImpl
| 281 - org.opendaylight.openflowplugin.impl - 0.4.0.SNAPSHOT | Flow add
failed for flow=AddFlowInput [_cookie=FlowCookie [_value=18],
_cookieMask=FlowCookie [_value=255], _flowName=FooXf18, _flowRef=FlowRef
[_value=KeyedInstanceIdentifier{targetType=interface
...
_priority=18, _tableId=2, _installHw=false, _strict=false, augmentation=[]],
errors=Device reported error type BADMATCH code BADWILDCARDS
Are you sure the problem is not in different ovs versions?
________________________________
From: Luis Gomez <[email protected]>
Sent: Monday, October 3, 2016 10:55 PM
To: Jamo Luhrsen; Miroslav Macko
Cc: [email protected];
[email protected]
Subject: Re: [integration-dev] [openflowplugin-dev] Failing robot tests
openflowplugin-csit-1node-flow-services
So looking at the FRS test in detail, the reason for the multiple failures is
not the OVS version but the FRS feature requiring more time to install flows +
update flow stats in oper DS.
In general it is bad practice to use sleep in tests but in this case it was
useful to flag the slowness of FRS, in particular:
- Flows_OF13 suite is failing because we give 1 sec between programming the
flow and see the flow in OVS, with FRS we need more time and that is why it is
fails.
- Stats_Manager_extended suite is failing because we give 3 secs between
programming flows and see them in operational DS, with FRS we need more time
and thats is why it fails.
- The Groups_Meters_OF13 is also failing because CPqD switch cannot connect to
controller, this could be a CPqD issue, with the old FRM is working but it
takes long time to connect.
Finally you can observe the differences in performance in these jobs:
https://jenkins.opendaylight.org/releng/view/openflowplugin/job/openflowplugin-csit-1node-periodic-scale-stats-collection-daily-frs-only-boron/plot/
https://jenkins.opendaylight.org/releng/view/openflowplugin/job/openflowplugin-csit-1node-periodic-scale-stats-collection-daily-only-boron/plot/
The plot titles are not very good but I can explain:
- Config Performance measures time for updating all flow stats in oper DS after
adding and after deleting.
- Performance rate measures rate and time for getting flows in config DS.
In both cases the FRM is superior than FRS.
BR/Luis
On Oct 3, 2016, at 9:20 AM, Jamo Luhrsen
<[email protected]<mailto:[email protected]>> wrote:
Miro,
both of the jobs you pointed to are using the same VM to run OVS.
"ubuntu-trusty-mininet-2c-2g"
is there something about the jobs that would make ovs have a different version?
As for using ovs 2.5, I recall hearing the plan would be to get all of the
openflowplugin csit
running 2.5+. I think it will create some issues that will need to be debugged
and fixed in our
CSIT.
JamO
On 10/03/2016 06:33 AM, Miroslav Macko wrote:
?Hi guys,
Could you please check this two robot tests [1] and [2]?
[1] is not failing
[2] is failing - some flows are rejected by device ("match inconsistent").
It looks like that [1] is running on the OVS version 2.0 and [2] is probably
running on some higher version of the OVS.
When we tried to use OVS 2.4 for [1], it was also failing with "match
inconsistent".
When we tried to use OVS 2.0 for [2], it is also not failing.
?
We have discussed it with Luis already, and he would like to use OVS version
2.5.0 for all tests.
So it will be needed to update failing flows xml input files
(/test/csit/variables/xmls).
Thank you.
Miro
[1]
https://jenkins.opendaylight.org/releng/view/openflowplugin/job/openflowplugin-csit-1node-flow-services-only-carbon/
[pastedImage.png]
[2]
https://jenkins.opendaylight.org/releng/view/openflowplugin/job/openflowplugin-csit-1node-flow-services-frs-only-carbon/
[pastedImage.png]
MiroslavMacko
Software Developer
Sídlo / Mlynské Nivy 56 / 821 05 Bratislava / Slovakia
R&D centrum / Janka Krála 9 / 974 01 Banská Bystrica / Slovakia
/ [email protected]<mailto:[email protected]>
reception: +421 2 206 65 114 / www.pantheon.sk<http://www.pantheon.sk/>
logo
_______________________________________________
openflowplugin-dev mailing list
[email protected]<mailto:[email protected]>
https://lists.opendaylight.org/mailman/listinfo/openflowplugin-dev
_______________________________________________
integration-dev mailing list
[email protected]<mailto:[email protected]>
https://lists.opendaylight.org/mailman/listinfo/integration-dev
AndrejLeitner
Software Developer
Sídlo / Mlynské Nivy 56 / 821 05 Bratislava / Slovakia
R&D centrum / Janka Krála 9 / 974 01 Banská Bystrica / Slovakia
/ [email protected]
reception: +421 2 206 65 114 / www.pantheon.sk
[logo]
_______________________________________________
openflowplugin-dev mailing list
[email protected]
https://lists.opendaylight.org/mailman/listinfo/openflowplugin-dev
