SCTP Discussions

Text archives Help


Re: [SCTP-Discussion] SCTP Path selection !!


Chronological Thread 
  • From: Michael Tuexen <Michael.Tuexen AT micmac.franken.de>
  • To: manju reddy <manju.reddy001 AT gmail.com>
  • Cc: sctp-discussion AT sctp.de
  • Subject: Re: [SCTP-Discussion] SCTP Path selection !!
  • Date: Thu, 4 Apr 2013 13:04:27 +0200
  • List-archive: <http://lists.franken.de/pipermail/sctp-discussion>
  • List-id: SCTP Discussions <sctp-discussion.sctp.de>

On Apr 4, 2013, at 4:49 AM, manju reddy wrote:

> Hi,
>
> I'm new to SCTP, so wanted to understand the SCTP behavior in case of
> Path selection.
>
> Hope this mailing list helps me here.
>
> I have an single homed server(running on solaris10 u8) and Multi-homed
> client(two-IP's).
> Assume client's IP is X and Server is using(Y,Z).
I guess you mean the server is multi homed, the client is single homed...
>
> In client side i have configured the PathMaxRetrans as 3, RtoMin as 500
> msec, RToMax as 2000, RtoInitial as 2000.
>
> At the start of the association the primary path selected was X -> Y and
> everything was fine. But due to some
This is the perspective of the client.
> network issues or some delays, i could see SACK for only packet from Y was
> delayed, which we re-transmitted after 500 msec to Z.
Timer based retransmissions for packets originally sent to Y would go to Z.
This makes sense.
>
> I hope that was fine till then, but after we started sending only to Z.
> But still due to different values and
OK. Possibly (don't know if Solaris implemented it) the path to Y is
considered partially failed and
therefore you are sending to Z.
> setting on server side, they are still sending the DATA from Y only to X.
This now is the reverse direction. Directions are independent.
>
> Now It has become something like this.
>
> X---->DATA--->Z
> X<---SACK<---Z
>
> X<---DATA<---Y
> X--->SACK--->Y
>
This is fine.
>
> But my question is even though the PathMaxRetrans as 3, why the path
> switched just for one failure??
Please note that this list is for supporting the sctplib implementation. I
think
you are asking Solaris specific questions... Possibly they have implemented
partially failed state.
>
> And when will it switch back to primary path??
Possibly once the path towards Y has no failures anymore. At least in the
FreeBSD kernel
implementation we don't switch primary paths automatically.
>
> Is there any configuration parameter like avoiding primary path or not???
> which may help me??
Using the SCTP_STATUS socket option, you can figure out which is the current
primary path
and using the SCTP_SET_PEER_PRIMARY_ADDR socket option you can set it.

See http://tools.ietf.org/html/rfc6458 for details. I think Solaris supports
the socket options.

Best regards
Michael
>
>
> Thanks for your time,
> Thanks in advance,
>
> Manju
>
>
>
>
>
> _______________________________________________
> sctp-discussion mailing list
> sctp-discussion AT sctp.de
> http://www.sctp.de/mailman/listinfo/sctp-discussion





Archive powered by MHonArc 2.6.19+.

Top of page