compute() has been optimized and all old code is gone. We no longer need to compute trees for each cluster. Furthermore, the computation of broadcast ports has been streamlined. Other routing/topology refactoring has been done as well.

Next round of refactoring. Attempting to normalize the terminology used. Fixed some bugs that have been around for a while too.

unit tests fixed by fixing an archipelago bug. A cluster is a *strongly connected* group of switches. A switch that is weakly connected to a cluster but not strongly connected to any other cluster will exist in its own cluster. This weak link, if not an external/BDDP-discovered link, will not be used to try and compute paths. Thus, while we compute clusters, any link that is not within a cluster and is not an external link is stored for reference when computing archipelagos. Essentially, the external links which typically are between clusters get merged into a new set with these weak links to allow for proper computation of archipelagos.

Create wrapper class that takes an OFMessage and allows for direct use in other containers where we don't care about the XID of the message. For example, if we want to create a unit test with predictable results but the XIDs are different depending on unit test execution, we can use the wrapper class to guarantee consistency. In the main code, we can more easily compare OFMessages without care for the XID and create cool data structures like caches.

Remove old makefile. Add round #1 of version support via REST API. The included code adds a /wm/core/version/json REST API. This will work in Eclipse and in situations where the jar is being run in the build environment with the pom.xml. It will not work when the jar has been compiled and relocated outside of the build environment. For this, we need to come up with a solution to build in the version at compile-time. This will either incorporate it as code or save it somewhere in the classpath.

Add in old URIs for static entry pusher for backwards compatibility. Note this only addresses URI compatibility *not* JSON key:value compatibility. Clean up warnings in LoadBalancer unit test

Finally! :-) OFMessageDamper works. Required creating a custom hashCode() (hashCodeIgnoreXid()) function for all OFMessages we would like to dampen.

This is a huge commit. Lots of (good) stuff has been added and updated. First and foremost, OpenFlow 1.5 is now supported (yay) via a major update to OpenFlowJ-Loxi. This update necessitated the update of the Static Flow Pusher module and MatchUtils, ActionUtils, and InstructionUtils to support the newly-added features. This prompted me to explore the SFP to see if there was anything that could be improved. First thing that came to mind was adding support to push OpenFlow groups in addition to flows, thus group support has been added to the SFP via JSON, which is the format that should have been used in the first place. Action lists will be updated in the near future to be JSON instead of one massive ',', '=', and '->' delimited string. Thus, the Static Flow Pusher has been renamed the StaticEntryPusher, since it now encompasses groups. While groups were being added to the SFP (ahem...SEP), a number of optimizations were made to the SEP code, greatly reducing its complexity (IMHO). The support for OpenFlow meters can now be added rather trivially in the future, if desired. Last thing to note about the SEP is the work-in-progress addition of a 'schema' API to allow northbound applications to get information about SEP syntax. This is not complete yet, but the skeleton code is in place. Next thing in this pull request is miscellaneous changes throughout the controller, primarily in (de)serializers and REST APIs to support OpenFlow 1.5. Next item of business is the removal of the DebugEventService. This is an artifact of days past and need not clutter the controller anymore. Same goes for the TestModule; now that we have better tutorials and documentation on the wiki, an example isn't necessary. And, last but not least, some old libraries that weren't being used have now been removed.

Updated to Java 8, junit-4.12, easymock-3.4. Still having trouble having unit tests pass in Eclipse for DeviceManagerImplTest, but they work find in maven and ant. testPacketInBasic and testPacketInBasicIPv6 are the trouble makers (fail occasionally).

Handle link and port down events in Forwarding. Remove flows ingress and egress a failed port or link to eliminate stale entries. This is done in Forwarding instead of the Link Discovery Manager, since some user modules might not want their flows removed and wish for them to persist until the link comes back up.

Hello guys, I have changed the logic of simple.FT (Fault Tolerant) module.

The basic idea is, the FT module register on SyncManager to receive RPC events
(connect and disconect nodes).

In connect events, the controller send its switch list to storage and
the other controllers can sync this info.
In disconnect events, the controller is informed about which controller crashed and
get switch list from storage and send a role request message to swicthes.

Regards.

and compare with initial cluster config defined in floodlightdefault.properties.

Issues: do not treat the message ROLE_STATUS.
In case a cluster goes down and goes up, there is a problem, the Controller does not change its own role to role slave.

But works properly as Primary-Backup configuration.

Under development...

 [junit] Running net.floodlightcontroller.devicemanager.internal.DeviceManagerImplTest
 [junit] Tests run: 37, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 0,503 sec
 [junit] TEST net.floodlightcontroller.devicemanager.internal.DeviceManagerImplTest FAILED

ant tests result in some failures.

SyncManager results:
13:08:12.696 ERROR [i.n.u.ResourceLeakDetector:main] LEAK: You are creating too many HashedWheelTimer instances.  HashedWheelTimer is a shared resource that must be reused across the JVM,so that only a few instances are created.

Include the completion listener in the MockFloodlightProvider

Added unit test for writing messages while in MASTER/SLAVE. Changed to a blacklist mechanism instead of a whitelist (to match how the spec describes invalid (write) SLAVE messages). Completed invalid SLAVE messages for each OF version.

Updated to properly handle IOFSwitch write()'s while in SLAVE role. TODO: make unit test for trapping invalid messages while in SLAVE. TODO: complete logic for handling valid SLAVE message types per OF version.

OMG, it was a timeout that was too short. Good news is that the bootstrap unit test passes. Bad news is that there is a timeout somewhere in the sync module that is way too long and causes the unit tests for sync to take literally a few minutes. Will look into that next.

Fixed a blunder in OFSwitchManager and Controller to allow switches to connect if a server IP isn't explicitly given in floodlihgtdefault.properties. Still can't get sync's bootstrap test to pass. Feels like a race condition b/t update threads.

Core ported to Netty 4 along with unit tests. Still need to do sync and some loxi-related compile errors in web.serializers.

Tidy up topology and get rid of some many redundant functions in ITopologyService. Also, fix a bug that prevents outgoing broadcast ports from being queried on single-switch islands.