Big Problem when a router is disconneted

Has anyone else experience this?

I have one coordinator sending packets to 3 routers every second. I have my network setup with One-To-Many routing with Ack disabled (frameid=0), and I’m using API firmware. What that means is that when the coordinator transmits a packet to a router, it only expects to receive a ‘route record indicator’ packet (which I treat in my code as an acknowledgment). Everything is great until I disconnect one of the routers, then communication to all other routers becomes way slow, taking several seconds (instead of less than the typical 100ms) to receive the “ack” packet.

To make it simple to visualize I recorded a video and posted on youtube. In the video, the coordinator is happily sending a packet to 3 routers and receiving a 'route record indicator’until I unplug one of the routers (at second 10 in the video). Then you can see towards the end that the communications gets really messy, acks take a long time.

http://www.youtube.com/watch?v=H1SqNTJMI2w

Is this a firmware bug (i’m using 218C) ??? or am I missing some configuration setting?

Unfortunately, I don’t think this is a bug. it probably relates to the XBee waiting for some hand-off despite you not wanting an ACK.

Bottom line, with ZigBee (aka: fw 2x8C) you have to assume worst case every packet will take up to 5 seconds to arrive at a remote. It is the nature of a “mesh” with robust auto-re-routing.

If you really want a deterministic radio (but is okay to LOSE mesh/routing), then the 802.15.4 allows broadcast and message delivery in the 20-40msec range … but no mesh-forwarding. It is all point-to-multi-point.

The problem is worse than that. When only 1 router is diconnected, the whole network goes to a crawl, the latency go from an average 100ms to up an average of 10000ms (10 seconds) for all nodes on the network and the number of packets successfully transmitted also becomes really bad (5 out of 6 fails).

I would be surprised if the standard allows for such thing and I’m concerned because this would be a show stopper for us to use Zigbee here in our company.

I’ve recorded another video on this. I remove router with address 634C from the network at second 13 in the video. Observe how the number of successful transmissions and latency (avg, max, etc) becomes really bad after that. I’ve also attached both my coordinator and router configurations just in case and would greatly appreciate help on this!

http://www.youtube.com/watch?v=9NdmPmSQsFY

I was told by Digi that this can be resolved by disabling Many-To-One routing.

If disconnecting a router causes the network to halt when Many-To-One routing is enabled, then this feature should probably never be enabled!