[CONTROLLER-905] Scale problem: connecting mininet with 127 or 256 switches locks up the controller Created: 28/Sep/14  Updated: 19/Oct/17  Resolved: 08/Oct/14

Status: Resolved
Project: controller
Component/s: mdsal
Affects Version/s: Helium
Fix Version/s: None

Type: Bug
Reporter: Jan Medved Assignee: Unassigned
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: Zip Archive log.zip    
External issue ID: 2106

 Description   

Attaching a mininet with 127 switches (when l2switch components installed) or 255 switches (when l2switch components not installed) causes the controller to locks up. Stats and inventory updates stop. The following exception is observed:

logoutException in thread "FlowCapableInventoryProvider" java.lang.IllegalStateException: Transaction chain has failed
at com.google.common.base.Preconditions.checkState(Preconditions.java:149)
at org.opendaylight.controller.md.sal.dom.broker.impl.DOMDataBrokerTransactionChainImpl.checkNotFailed(DOMDataBrokerTransactionChainImpl.java:80)
at org.opendaylight.controller.md.sal.dom.broker.impl.DOMDataBrokerTransactionChainImpl.submit(DOMDataBrokerTransactionChainImpl.java:91)
at org.opendaylight.controller.md.sal.dom.broker.impl.DOMForwardedWriteTransaction.submit(DOMForwardedWriteTransaction.java:142)
at org.opendaylight.controller.md.sal.binding.impl.AbstractWriteTransaction.doSubmit(AbstractWriteTransaction.java:142)
at org.opendaylight.controller.md.sal.binding.impl.BindingDataWriteTransactionImpl.submit(BindingDataWriteTransactionImpl.java:83)
at org.opendaylight.controller.md.sal.binding.impl.BindingTranslatedTransactionChain$1.submit(BindingTranslatedTransactionChain.java:68)
at org.opendaylight.controller.md.inventory.manager.FlowCapableInventoryProvider.run(FlowCapableInventoryProvider.java:124)
at java.lang.Thread.run(Thread.java:745)

The condition is very repeatable and has been reproduced on multiple systems.

The bug is raised against md-sal, although it's not an md-sal bug. It will likely migrate into one of md-sal clients when more thorough analysis has been done.

 Comments   
Comment by Jan Medved [ 28/Sep/14 ]

Attachment log.zip has been added with description: Logs from start until the controller locked up

Comment by Tony Tkacik [ 29/Sep/14 ]

The root cause is commit queue full, but not sure about which MD-SAL client polutes it.

Comment by Tony Tkacik [ 01/Oct/14 ]

https://git.opendaylight.org/gerrit/#/c/11663/

Please verify this bug is fixed and close the issue afterwards.

Generated at Wed Feb 07 19:54:10 UTC 2024 using Jira 8.20.10#820010-sha1:ace47f9899e9ee25d7157d59aa17ab06aee30d3d.