On Mar 14, 2011, at 3:36 AM, Will Yu wrote:

> Hello dear Michael,
> 
> Thank you for your reply. There is an additional query about your reply:
> In our realization, the remote address is regarded as unreachable until the 
> Heartbeat Message from TWO different local address(IP1 and IP2) exceed PMR. 
> It means that IP1 sent 4(PMR value) Heartbeat and IP2 sent 4(PMR value) 
> Heartbeat as well. In your describing, is this method not accord with RFC?
The RFC only talks about an error counter per remote address, not an error 
counter per local/remote address pair.

However, you should be able to handle a network setup, where no cross routing 
is possible,
so there is only connectivity between IP1 <-> IP3 and IP2 <-> IP4. This is a 
setup which
is often used in SIGTRAN setups.

Best regards
Michael
> 
> Thanks again.
> 
> Brs,
> Will Yu
> 
> 2011/3/12 Michael Tüxen <[email protected]>
> 
> On Mar 11, 2011, at 7:44 AM, Will Yu wrote:
> 
> > Hello,
> >
> > I have a query on SCTP standard (RFC 4960). In section 8.2 Path Failure 
> > Detection, it describes the action which end points should take in 
> > detecting path unavailable.
> >
> >    When its peer endpoint is multi-homed, an endpoint should keep an
> >    error counter for each of the destination transport addresses of the
> >    peer endpoint.
> >    Each time the T3-rtx timer expires on any address, or when a
> >    HEARTBEAT sent to an idle address is not acknowledged within an RTO,
> >    the error counter of that destination address will be incremented.
> >    When the value in the error counter exceeds the protocol parameter
> >    ’Path.Max.Retrans’ of that destination address, the endpoint should
> >    mark the destination transport address as inactive, and a
> >    notification SHOULD be sent to the upper layer.
> >
> > In my case, I connected two end points through four IP address. EP1 owns 
> > IP1 and IP2, EP2 owns IP3 and IP4. And IP1 is the primary address for EP1, 
> > while IP3 is the primary address for EP2. There is no other packet in this 
> > association except with Heartbeat and Heartbeat_ACK.
> >
> >                 IP1--------------------IP3
> >     End                   \ /                   End
> >    Point1                 *                   Point2
> >                              / \
> >                 IP2--------------------IP4
> >
> > When IP4 is unavailable, the heartbeat messages from IP1 and IP2 with the 
> > destination IP4 cannot received acknowledgement. In this case, PMR in each 
> > EP is set to 4. I observed that EP1 used IP1 and IP2 as source address to 
> > send the heartbeat messages to IP4 respectively. When the heartbeat message 
> > from source address IP2 exceed 4 times, it marked the path(IP2-IP4) is 
> > unreachable. And then IP1 exceed, it would mark the destination IP4 is 
> > unavailable. In my opinion, this result is
> There is no state for the path IP2-IP4, only for the remote address IP4.
> > reasonable with the recognition of IP2-IP4 is a path and IP1-IP4 is another 
> > path for the association. Consequently, the PMR parameter is for 
> > determining which PATH could regard as unavailable.
> SCTP supervises remote addresses. So EP1 is supervising IP3 and IP4. So there 
> is one  error counter for IP3
> and one for IP4. There is also only a state per remote IP address. So after 
> the error counter for IP4
> exceeds PMR, the IP address IP4 should be marked as unreachable.
> > But in the section 8.2 Path Failure Detection, the increasing counter 
> > action is based on different destination transport address, not on the 
> > different source-destination address pair. I think it make me confused that 
> > counter will impact the result in my case with the assumption RFC regard 
> > the destination address failure as the critical element for Path Failure 
> > Detection.
> >
> > Would you kindly help me clarify the behaviors in Path Failure Detection, 
> > or figure out if I have any misunderstanding in this case?
> As said above, SCTP supervises remote addresses, not paths.
> 
> Best regards
> Michael
> 
> PS: It might be better to send SCTP related questions to [email protected]
> >
> > Brs,
> > Will Yu
> > _______________________________________________
> > Ietf mailing list
> > [email protected]
> > https://www.ietf.org/mailman/listinfo/ietf
> 
> 

_______________________________________________
Ietf mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/ietf

Reply via email to