Node disconnect issue while polling

Darkscene

Other Parts Discussed in Thread: Z-STACK

I'm using the Z-stack Linux gateway for my home automation project. I'm polling nodes using the api - device_send_list_maintain_request () (nwk address request ) to update the nodes' activity in the network.Problem is that node's status gets updated for a certain period (about 1 hour) and after that,all the nodes get disconnected from the gateway.Polling happens once in every 50 seconds. The timeout is set to as 9s and maximum failed attempts is set as 5. Any suggestions are welcome.

over 10 years ago

0 YiKai Chen over 10 years ago

Guru 735685 points

When "all the nodes get disconnected from the gateway" happens, do you use Ubiqua Packet Analyzer or Packet Sniffer to check if ZEDs are lost connection with GW?

0 Darkscene over 10 years ago in reply to YiKai Chen

Prodigy 210 points

Hi Chen,
The nodes before getting disconnected becomes inactive. In that case control actions are possible from gateway to ZEDs and ZEDs send link status to gateway (as seen from Ubiqua packet capture). But after disconnecting from gateway no more control actions could be performed.

0 YiKai Chen over 10 years ago in reply to Darkscene

Guru 735685 points

The problem is why all of your ZEDs lose connection with coordinator.

0 Darkscene over 10 years ago in reply to YiKai Chen

Prodigy 210 points

yes and i need to know why all of the ZEDs lose connection with coordinator.

0 YiKai Chen over 10 years ago in reply to Darkscene

Guru 735685 points

That is why I ask you use Ubiqua Packet Analyzer or Packet Sniffer to check.

0 Darkscene over 10 years ago in reply to YiKai Chen

Prodigy 210 points

In ubiqua packet capture after disconnecting of the nodes, both nodes and gateway send link status. ZIGBEE_GENERIC_CNF Status shows SUCCESS and

GW ZIGBEE_GENERIC_RSP_IND Status shows FAILURE.

0 YiKai Chen over 10 years ago in reply to Darkscene

Guru 735685 points

I don't see the sniffer log attached. By the way, can you also provide the sniffer log before this happens?

0 Darkscene over 10 years ago in reply to YiKai Chen

Prodigy 210 points

im attaching the ubiqua packet log after the disconnect.ubiqua_aftr_disconnect.txt

0 YiKai Chen over 10 years ago in reply to Darkscene

Guru 735685 points

I have ubiqua packet analyzer so can you attach sniffer log as *.cubx?

0 Darkscene over 10 years ago in reply to YiKai Chen

Prodigy 210 points

ubiqua packet capture file .ubiq_node_offline.cubx.tar.gz

0 YiKai Chen over 10 years ago in reply to Darkscene

Guru 735685 points

In your sniffer log, I see there are three different Zigbee network and there is no end device. All are coordinator and router. Are you sure you are testing on end devices?

0 Darkscene over 10 years ago in reply to YiKai Chen

Prodigy 210 points

That is because I have configured all nodes as routers. (For the Zigbee mesh networking)

0 YiKai Chen over 10 years ago in reply to Darkscene

Guru 735685 points

ZR is not like ZED which would do polling to its parent node to know if it loses connection with its parent node. ZR won't have this behavior. I am confused by you now. What is your question and expectation?

0 Darkscene over 10 years ago in reply to YiKai Chen

Prodigy 210 points

As per my implementation, there are no ZEDs in the network . Gateway is coordinator and all nodes are routers. Gateway polls the nodes periodically every 50 seconds with device_send_list_maintain_request. This is required as the user may manually power off the nodes and at that time we need to mark the node status as offline at gateway software. And later when he turns it on we need to mark it as online at gateway software

0 YiKai Chen over 10 years ago in reply to Darkscene

Guru 735685 points

So, what is your problem?

0 Darkscene over 10 years ago in reply to YiKai Chen

Prodigy 210 points

For the first 15-30 minutes the polling is working fine and Zigbee servers mark the nodes as offline if I manually powered it off and mark as online when it is turned on. But after 15-30 minutes we are observing the nodes are going offline from the network. Even though the nodes are marked as offline, we are still able to send Zigbee control signals to the node for a few more minutes, but with a very less success rate. We have tested this with our designed nodes and TI ZLL light node also, but the issue there in all the nodes. We are still able to see Zigbee packets from the coordinator and routers even after this issue happens.

0 YiKai Chen over 10 years ago in reply to Darkscene

Guru 735685 points

What do you send in device_send_list_maintain_request?

0 Darkscene over 10 years ago in reply to YiKai Chen

Prodigy 210 points

In device_send_list_maintain_request, ieee_addr and endpoint of each nodes are send and it is called per node basis.The device_send_list_maintain_request in turn calls MT_ZDO_NWK_ADDR_REQ API.

0 YiKai Chen over 10 years ago in reply to Darkscene

Guru 735685 points

Do you mean that your ZC can get MT_ZDO_NWK_ADDR_RSP after you run your application for a while?

0 YiKai Chen over 10 years ago in reply to YiKai Chen

Guru 735685 points

If it is possible, please attach the full Ubiqua sniffer log including this problem happens.

0 Darkscene over 10 years ago in reply to YiKai Chen

Prodigy 210 points

Hi Chen,

Please refer the ubiqua capture files which include the nodes before going offline(capture 1-4) , going offline(capture 5),after going offline (capture 6).

7673.captures.zip

0 YiKai Chen over 10 years ago in reply to Darkscene

Guru 735685 points

In the sniffer log, I see you send NWK address request continuously and I believe this might be the problem. I would suggest you try to send NWK request with at least 100 milli-seconds delay in between two requests.

0 Darkscene over 10 years ago in reply to YiKai Chen

Prodigy 210 points

I am polling each ZR node using NWK address request with one second delay. And this polling is repeated after an interval of 50 seconds.

0 YiKai Chen over 10 years ago in reply to Darkscene

Guru 735685 points

You can check your Ubiqua log and will see NWK address requests are sent in 3 to 5 milli-second interval which I think it is too fast.

Zigbee & Thread

Zigbee & Thread forum

Node disconnect issue while polling