As it turns out, the root of the problem was caused by one of the Cisco dudes fat-fingering a config on an uplink. All six servers in two datacenters are quite happily awaiting our DTU deployment! We utilize multicast quite substantially in our production applications, so it surprised me that this wasn't working out of the gate.
This being an "unsupported configuration" begs the question, though... Is there a better way to do this? In order to support my SLA on the DTU's, I need to share my group across datacenters. We don't route VLANs across datacenters, which only leaves multicast routing. It seems like a very elegant solution, so I'm just curious... What does Sun do? Thanks again, you guys are a *very* valuable asset! -Bill -----Original Message----- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of ottomeister Sent: Saturday, September 01, 2007 2:37 PM To: SunRay-Users mailing list Subject: Re: [SunRay-Users] Multicast problems On 8/31/07, Quayle, Bill (ETS) <[EMAIL PROTECTED]> wrote: > I've been working on the deployment of a Sun Ray server installation across > two datacenters. All networking is Cisco gear, and PIM is enabled for the > VLANs on all switches and routers in the path. There are 6 hops from server > to server, so I have set the TTL in auth.props to 10 (compensating for any > future network changes.) The multicast IP is private and not conflicting. The fact that the servers don't have a subnet in common means that this configuration is officially unsupported. The subnet-in-common requirement exists precisely because multicast across subnet boundaries is still, annoyingly, somewhat of a crapshoot in the real world. It *should* work, but often it doesn't. > I've captured outgoing multicast packets from the servers, and have found > that the TTL of the IGMP packets is still set to 1. (I'm not sure if this > is a problem.) Not a problem. IGMP is for the consumption of on-subnet machines. It would be a bug if the IGMP TTL was anything other than 1. > The UDP packets show the proper TTL of 10. I can provide a > sample of the packets, if needed. I snoop the interface at one datacenter > and don't see any multicast packets (UDP or IGMP) originating from the > servers at the other site, and vice-versa. I've verified the group > signatures as being in sync across the board. The utgstatus command confirms > that the group is restricted to only the servers in a single datacenter. > > I need to get a group working across datacenters to support our SLA for DTU > availability. Any help or advice would be appreciated. The first thing I'd check would be that the routers understand IGMP version 3. That's what the Solaris 10 machines will be sending. If the router firmware is downrev then it might only understand IGMPv2 and v1. The routers are probably maintaining some sort of IGMP statistics so one quick way to check is to examine those and see whether they mention IGMPv3 at all. OttoM. __ ottomeister Disclaimer: These are my opinions. I do not speak for my employer. _______________________________________________ SunRay-Users mailing list [email protected] http://www.filibeto.org/mailman/listinfo/sunray-users _______________________________________________ SunRay-Users mailing list [email protected] http://www.filibeto.org/mailman/listinfo/sunray-users
