[OPNFLWPLUG-522] [GROUP RECONCILIATION] Group Chaining fails during reconciliation Created: 30/Jul/15  Updated: 27/Sep/21  Resolved: 30/Jan/16

Status: Resolved
Project: OpenFlowPlugin
Component/s: General
Affects Version/s: None
Fix Version/s: None

Type: Bug
Reporter: Pompina Singh Assignee: Shuva Jyoti Kar
Resolution: Done Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified
Environment:

Operating System: Linux
Platform: All

Attachments: File Group-Chaining-issue.pcap    
External issue ID: 4069
Priority: Normal

 Description   

Description:

1. After creating the setup, connected a cpqd switch to the controller.
2. Created a group-chaining scenario in which an all group is pointing to 2 indirect groups as in

stat_repl{type="gdesc", flags="0x0", stats=[
{type="all", group="1", buckets=[{w="0", wprt="255", wgrp="2", acts=[group

{id="3"}

]}, {w="0", wprt="255", wgrp="2", acts=[group

{id="2"}

]}]},
{type="ind", group="2", buckets=[

{w="0", wprt="0", wgrp="0", acts=[vlan_pop]}]},
{type="ind", group="3", buckets=[{w="0", wprt="0", wgrp="0", acts=[vlan_pop]}

]}]}

3. Restart the switch 2/3 times , we are left with only the children. The parent is not present as in

stat_repl{type="gdesc", flags="0x0", stats=[
{type="ind", group="2", buckets=[

{w="0", wprt="0", wgrp="0", acts=[vlan_pop]}]},
{type="ind", group="3", buckets=[{w="0", wprt="0", wgrp="0", acts=[vlan_pop]}

]}]}

4.The configuration and the operational datastore thus have different results.

This situation is even more worse when we have flows pointing to groups as the following flow entry that would point to the parent group entry would also fail.

This might result in traffic drop under normal scenarios, hence there's a problem during reconciliation.

 Comments   
Comment by Pompina Singh [ 30/Jul/15 ]

Attachment Group-Chaining-issue.pcap has been added with description: Attached Pcap shows the problem

Comment by Hariharan Sethuraman [ 07/Aug/15 ]

Hi Pompina,

When you say config-ds and oper-ds are out of sync: config-ds had only 2 groups and oper-ds had 3 or the other way?

This will help us to look from where to begin the debugging.

Thanks,
Hari

Comment by Shuva Jyoti Kar [ 08/Aug/15 ]

<Replying on behalf of Pompina>

Hi Hari

Yes the config ds has 3 groups while the operational has 2 groups. This is because the provisioning of the top-level(parent/dependent) group (here in this case the one with group-id 1 has failed ) while provisioning of the children/independent groups (here with group ids 2 and 3 ) succeeded.

In this case if there are flows pointing to the dependent group that would fail also resulting in traffic loss during reconciliation

Comment by Hariharan Sethuraman [ 14/Sep/15 ]

Shuva is already working in the reconciliation and working on review comments. Requested him to take this defect and 4099.Setting the owner to default.

Comment by Abhijit Kumbhare [ 25/Sep/15 ]

Assigning it to Shuva since he is already working on this as part of reconciliation

Comment by Abhijit Kumbhare [ 09/Oct/15 ]

Any thoughts/update Shuva?

Comment by Shuva Jyoti Kar [ 09/Oct/15 ]

https://git.opendaylight.org/gerrit/#/c/27880/

Build failing at openflowplugin-it...shared it with integration team...it seems to be an open issue still

Comment by Shuva Jyoti Kar [ 19/Oct/15 ]

Changes merged to the mainline post incorporating review comments

Comment by Abhijit Kumbhare [ 30/Oct/15 ]

Shuva,

Luis thinks this is no longer an issue - can you confirm and close it?

Thx,
Abhijit

Comment by Luis Gomez [ 30/Oct/15 ]

To be precise, I said I do not think there is issues in the patch build as pointed out but still this patch has to be tested and verified for the bug.

Comment by Shuva Jyoti Kar [ 02/Nov/15 ]

Luis, I meant the intermediate build failed few times, post that it was merged once the build went through, am i missing something
Also , do you require validation of the fix in terms of test-cases ?

Comment by Luis Gomez [ 02/Nov/15 ]

To close the bug it is enough with manual validation, but as I commented to Abhijit it would be nice if Ericsson could build some system test around reconciliation so we do not regress this key feature.

Comment by Muthukumaran Kothandaraman [ 03/Nov/15 ]

Hi Luis,

We started working on creating integration tests around reconciliation. We are reviewing the same and would be submitting for review sooner. Once approved, these TCs can be mainstreamed for regular integration runs

Hope this is fine

Comment by Luis Gomez [ 03/Nov/15 ]

That sounds really good, waiting forward to seeing your patches in integration/test.

Comment by Abhijit Kumbhare [ 01/Dec/15 ]

Needs to be cherry picked to stable/lithium?

Comment by Abhijit Kumbhare [ 25/Jan/16 ]

Shuva,

Isn't this merged? i.e. https://git.opendaylight.org/gerrit/#/c/27880/. Should we close this bug?

Thanks,
Abhijit

Comment by Abhijit Kumbhare [ 30/Jan/16 ]

https://git.opendaylight.org/gerrit/#/c/27880/

Generated at Wed Feb 07 20:32:42 UTC 2024 using Jira 8.20.10#820010-sha1:ace47f9899e9ee25d7157d59aa17ab06aee30d3d.