[OVSDB-286] ovs node removed from operational after partitioned from network and a config is made Created: 02/Feb/16  Updated: 28/Jun/16  Resolved: 28/Jun/16

Status: Resolved
Project: ovsdb
Component/s: Southbound.Open_vSwitch
Affects Version/s: unspecified
Fix Version/s: None

Type: Bug
Reporter: Jamo Luhrsen Assignee: Ananthi Palaniswamy
Resolution: Done Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified
Environment:

Operating System: All
Platform: All

Attachments: Text File 5178.karaf.log     Text File Steps_toreproduce.txt     Text File karaf.log     File karaf.log-1    
External issue ID: 5178

 Description   

the final result is an ovs instance thinking it's connected to the plugin (e.g. "is_connected: true") but it does not exist in operational.

steps to reproduce:

1. connect ovs node to plugin (ovs-vsctl set-manager)

2. partition node from controller with iptables:
sudo /sbin/iptables -A OUTPUT -p tcp --dport 6640 -j DROP

3. clear ip tables so communication is ok again.

4. create a bridge via northbound rest api

notes:

It doesn't appear that the plugin times out the connected node when it
stops hearing from it (between steps 2 and 3). I've let it sit for many
minutes like this and it remains in operational.

after step 3, the plugin throws an exception because the node connecting
again "was already registered for ownership"

after step 4, the plugin removes the node because it "has no owner, cleaning up the operational data store"

full karaf log attached

 Comments   
Comment by Jamo Luhrsen [ 02/Feb/16 ]

Attachment 5178.karaf.log has been added with description: karaf.log

Comment by Jamo Luhrsen [ 06/Apr/16 ]

Attachment karaf.log-1 has been added with description: another newer log for reference

Comment by Jamo Luhrsen [ 06/Apr/16 ]

This is still seen in CSIT. There is not any of the "partitioning" happening
either, like I initially wrote. Through the normal course of running suites
this is hit. I attached a recent karaf.log from ovsdb-csit-1node-southbound-all-beryllium.

It's reproducible locally as well, just running some of the ovsdb suites
back to back. You can get in a state where everything appears clean, and
then try to connect ovs to southbound and we end up in this bug state.

by clean, I mean there is nothing in config or operational store. ovs
is showing not connected. Simply doing vsctl set-manager and ovs shows
that it is connected, then ODL operational does not have it.

in karaf.log you can see this message when this happens:

2016-04-06 21:23:19,854 | INFO | lt-dispatcher-17 | OvsdbConnectionManager | 173 - org.opendaylight.ovsdb.southbound-impl - 1.2.3.beryllium-daily-v201604020031 | Entity{type='ovsdb', id=/(urn:TBD:params:xml:ns:yang:network-topology?revision=2013-10-21)network-topology/topology/topology[

{(urn:TBD:params:xml:ns:yang:network-topology?revision=2013-10-21)topology-id=ovsdb:1}

]/node/node[

{(urn:TBD:params:xml:ns:yang:network-topology?revision=2013-10-21)node-id=ovsdb://209.132.179.50:6634}

]} has no owner, cleaning up the operational data store

Comment by Ananthi Palaniswamy [ 21/Apr/16 ]

Attachment Steps_toreproduce.txt has been added with description: Steps followed to reproduce the bug

Comment by Ananthi Palaniswamy [ 21/Apr/16 ]

Hi Jamo,

Attached the steps followed , not getting the exception as you have mentioned in attached karaf log. Reproduced the bug used latest Beryllium and attached recent karaf log.

Comment by Ananthi Palaniswamy [ 21/Apr/16 ]

Attachment karaf.log has been added with description: Attached recent karaf log

Comment by Anil Vishnoi [ 28/Jun/16 ]

Hi Jamo, Looks like this issue is fixed now. I am closing this bug, but if you still see this bug, please re-open it.

Generated at Wed Feb 07 20:35:59 UTC 2024 using Jira 8.20.10#820010-sha1:ace47f9899e9ee25d7157d59aa17ab06aee30d3d.