Fix for issue 685 #690
Fix for issue 685 #690
Conversation
Signed-off-by: Dscano <d.scano89@gmail.com>
Signed-off-by: Dscano <d.scano89@gmail.com>
| def deleteFlowRule(notif): | ||
| notif["sw"].DeleteTableEntry(notif["idle"].table_entry[0]) | ||
| table_entry = global_data['p4info_helper'].buildTableEntry( |
There was a problem hiding this comment.
This is just a guess at this point, but when I tried this and saw error messages, I added some debug print statements to deleteFlowRule and addFlowRule that showed all of these things:
(a) the switch name
(b) the protocol
(c) source IP address
(d) dest IP address
I saw that it seemed to be trying to delete the same rule (with the same key fields) from the same switch multiple times. It should definitely cause some kind of error to be returned from the network device to the controller if the controller attempts to delete an entry that is not in the table, but I do not know if that is the root cause of the error I am seeing.
If this is what is happening, I am also not yet certain why. My best guess would be that there is some kind of a race where the controller receives the first entry timeout notification, but before the switch receives the first delete command, the switch sends a second entry timeout notification for the same entry. Since such races are probably unavoidable in general, it seems like it would be a good idea for the controller to somehow correctly process error responses for attempted entry delete commands, in case they occur.
There was a problem hiding this comment.
I proposed a fix, as the multiple deletions of the same flow rule cause the gRPC error.
…le deletions Signed-off-by: Dscano <d.scano89@gmail.com>
|
Once the proposed code is approved, I need to update the README. |
This code fix the #685.