On 5/27/2010 at 04:18 AM, Alan Jones <[email protected]> wrote:
> Is there any interest in checking the on-the-wire version of the corosync
> protocol?
> For example, if the format of corosync messages changed in a way that was
> not backward compatible it might be nice to detect it up front.
Semi-related: it would be really nice if there were a way to unambiguously
detect corosync packets (like, a signature or something that's visible
even if secauth is on). ATM, AFAICT, all you can do is:
1) Assume anything on port 5405 is corosync.
2) If someone has chosen to run on a different port, look for multicast
traffic whose source port == destination port + 1, and hope it's not
some other protocol with the same semantics.
Unfortunately, unless there actually is some sort of signature I'm not
aware of, this presumably requires a change to the on-the-wire protocol,
and I'm guessing nobody will be happy about that idea.
Regards,
Tim
--
Tim Serong <[email protected]>
Senior Clustering Engineer, OPS Engineering, Novell Inc.
_______________________________________________
Openais mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/openais
