[mpls-tp] Comments on draft-ietf-mpls-tp-linear-protection-06.txt

[Lavanya Srivatsa <[e-mail]>]

To the authors,

I have a list of protection switching scenarios that do not seem to result in expected behaviour if operating as per this draft. I have proposed solutions/alternate text. I would appreciate your comments/confirmation on the same.

Scenario 1
Assume that there is a revertive protection switching group set up between routers R1 and R5 for working LSP1 and protection LSP2

[Sequence of events]
Both R1 and R5 in Normal state --> SF on R1 --> FS on R5 --> FS clear on R5
[Result]
Both R1 and R5 going to Normal state with NR(0,0)
[Problem]
The above is an incorrect state on R1 since the local Signal Fail still exists and has not been cleared. So R1 should have moved back to the local Protecting Failure state transmitting SF(1,1) instead of going to the NR(0,0) state.
[Suggested Solution]
In Section 4.3.3.3, the 2nd last bullet item under remote messages needs to be modified as - "A remote NR(0,0) message SHALL be ignored if in local Protecting administrative state.  If in remote Protecting administrative state then the LER SHALL go to Normal state and begin transmitting a NR(0,0) message OR shall go to the local Protecting Failure state if local Signal Failure is still reasserted and begin transmitting SF(1,1).

Scenario 2
Assume that there is a non-revertive protection switching set up between routers R1 and R5 for working LSP1 and protection LSP2.

[Sequence of events]
Both R1 and R5 in Normal state --> SF on R1 --> SF clear on R1 --> FS on R1 --> FS clear on R1
[Result]
Both R1 and R5 going to Normal state with NR(0,0)
[Problem]
Upon clearing the Forced Switch at R1, it is expected to go local Do-Not-Revert state again and begin transmitting DNR(0,1). Upon receiving this R5 is expected to go to remote Do-Not-Revert state.
[Suggested Solution]
In Section 4.3.3.3, the 1st bullet item under local input needs to be modified as - "A local Clear SHOULD be ignored if in remote Protecting administrative state.  If in local Protecting administrative state then this input SHALL cause the LER to go into Normal state and begin transmitting a NR(0,0) message if in revertive mode or go into Do-Not-Revert state and begin transmitting DNR(0,1) if in non-revertive mode.

Scenario 3
Assume that there is a protection switching set up between routers R1 and R5 for working LSP1 and protection LSP2 where R1 is operating in the revertive mode and R5 is operating in the non-revertive mode.

[Sequence of events]
Both R1 and R5 in Normal state --> SF on R1 --> SF clear on R1
[Result]
R5 moves to Wait-to-restore state and begins transmitting NR(0,1)
[Problem]
R5 is in non-revertive mode and as such does not have a Wait-to-restore state.
[Suggested Solution]
In Section 4.3.3.4, the 4th bullet under remote messages needs to be modified as - "If in remote Protecting failure state, a remote Wait-to-Restore message SHALL cause the LER to go into remote Wait-to-Restore state if in revertive mode and into remote Do-Not-Revert state if in non-revertive mode and continue transmission of the current message.


I have worked out the exact sequence details of the state machine for the scenarios above, which I have listed below.

- Lavanya


DETAILED SEQUENCE

Scenario 1
Assume that there is a revertive protection switching group set up between routers R1 and R5 for working LSP1 and protection LSP2.
Intially both are in Normal state sending NR(0,0) to each other.
Now if R1 detects a Signal Failure, R1 moves to the local Protecting Failure state and sends SF(1,1) to R5.
R5, on receiving SF(1,1), now moves to the remote Protecting Failure state and will transmit NR(0,1) to R1.
R1, on receiving NR(0,1), will ignore the message.
Now issue a Forced Switch command on R5.
R5 will now move to the local Protecting Administrative state and start transmitting FS(1,1).
R1, on receiving FS(1,1), now moves to remote Protecting Adminstrative state and since it was earlier in local Protecting Failure state, it will now transmit SF(1,1) to R5.
R5, on receiving SF(1,1) from R1, will ignore the message since there is an active Local Forced Switch command.
Now issue a Clear command on R5 for clearing the Forced Switch.
R5 will now move to the Normal state since it was in local Protecting Adminstrative state and start transmitting NR(0,0).
R1, on receiving NR(0,0), will go to Normal state since it was in remote Protecting Adminstrative state and begin transmitting NR(0,0).


Scenario 2
Assume that there is a non-revertive protection switching set up between routers R1 and R5 for working LSP1 and protection LSP2.
Intially both are in Normal state sending NR(0,0) to each other.
Now if R1 detects a Signal Failure, R1 moves to the local Protecting Failure state and sends SF(1,1) to R5.
R5, on receiving SF(1,1), now moves to the remote Protecting Failure state and will transmit NR(0,1) to R1.
R1, on receiving NR(0,1), will ignore the message.
Now SF clears on R1.
R1 will go to Do-Not-Revert state and start transmitting DNR(0,1).
R5, on receving DNR(0,1) moves to the remote Do-Not-Revert state and continues transmitting current message of NR(0,1).
R1, on receiving NR(0,1), will ignore this message.
Now issue a Forced Switch command on R1.
R1 will go the local Protecting Administrative state and begins transmitting FS(1,1) to R5.
R5, on receiving FS(1,1), moves to the remote Protecting Administrative state and begins transmitting NR(0,1).
R1, on receiving NR(0,1) will ignore this message.
Now issue a Clear Forced Switch command on R1.
R1 moves to the Normal state and begins transmitting NR(0,0).
R5, on receiving NR(0,0), moves to the Normal state and begins transmitting NR(0,0).


Scenario 3
Assume that there is a protection switching set up between routers R1 and R5 for working LSP1 and protection LSP2 where R1 is operating in the revertive mode and R5 is operating in the non-revertive mode.
Intially both are in Normal state sending NR(0,0) to each other.
Now if R1 detects a Signal Failure, R1 moves to the local Protecting Failure state and sends SF(1,1) to R5.
R5, on receiving SF(1,1), now moves to the remote Protecting Failure state and will transmit NR(0,1) to R1.
R1, on receiving NR(0,1), will ignore the message.
Now SF clears on R1.
R1 will go to Wait-To-Restore state, start the WTR timer and start transmitting WTR(0,1).
R5, on receiving WTR(0,1), moves to the remote Wait-To-Restore state and continues transmission of current message NR(0,1).