12-31-2018 11:20 AM - edited 12-31-2018 11:33 AM
I have the following configuration:
AF-5 bridge with an uplink to a Cisco 2960 Catalyst switch on the CO and CPE sides.
On the CPE end of the AF bridge, I have a Rocket 2AC bridge connected to the same 2960.
The CPE side of the Rocket 2AC bridge is connected to a Catalyst 2960. On that same 2960, I have a Rocket Prism5AC Gen2 sector. That sector has 3 CPE's. Two of the CPE's do not connect to a switch, but one does connect to a Cisco 2960.
The entire network functions reliably except for the CPE end of the Rocket 2AC bridge. On occasion, the Rocket Prism 5AC sector and all CPE radios connected to that sector will stop responding. The 2960 is usually accessible during these outages, and one thing I have noted during the outage is that the MAC table for the downlink to the 5AC sector is not populated, but the link state is UP.
I do have a couple devices connected to the same Cisco 2960 that the 5AC sector is connected to, and those devices always continue to respond during these "outages".
At this time last year, I had completely different Ubiquiti radios (apart from the AirFiber) at these locations, and we had the same random outages. We ended up putting dumb switches at those two locations, and everything operated much, much better and we were free from random outages.
We have tried trunk ports and access ports on our 2960's, neither seems to make the problem any better. We have also swapped the switches out with new switches.
All radios are on 8.5.8. I've tried rebooting everything (that is accessible) during these outages, doesn't seem to help. It all just comes back online after an inordinate amount of time. There also seems to be no consistency to when the outages happen. We will be fine for a couple weeks, then at some random time of day, an outage will occur. It may come back after 60 seconds, or it may take 10 minutes or more. It may happen once or twice, or more.
I suspect some sort of BPDU or ST event is happening, or maybe it's something to do with the MAC table on the CPE end of the sector, but I haven't verified that with either the Cisco or Ubiquiti side of things.
We have several hundred nodes on our private nework including other Ubiquiti bridges and links that operate without error all year long. It just seems to be this last-last mile bridge to sector to CPE with a Cisco 2960 that causes a problem.
I could go back to dumb switches, but that would just be... dumb. It's all just L2 out there.
Thank you in advance for your assistance.
01-02-2019 09:59 AM
Do you have a simple network diagram?
Do you have spanning tree enabled on the APs?
Do you lose access via ethernet to the AP, but not the switch it is connected to?
Any indicators in the logs of the switch?
01-02-2019 10:29 AM
Attached is a network design. Items outlined in dasehd red are what goes down.
Spanning Tree is not enabled on the APs
We do lose access to the 5AC AP (and everyting downstream from the 5AC sector of course) but not the switch.
Switch doesn't have anything like err-disabled or that there was some abnormal event. We are syslogging events >=6
01-02-2019 10:38 AM
Does the ethernet link go down as well (from to R5AC PRISM)?
Could you forward the Support Info file (System tab) from the AP and Station that is connected to a switch? Please send to firstname.lastname@example.org and include a reference to this thread.
01-02-2019 11:57 AM - edited 01-02-2019 12:05 PM
When the symptom is happening, the link light on the switch will go out for several seconds, then it will come on and the AP IP address will respond for about 9 or 10 pings, then the light goes and and we start all over again.
Yes, I will send those files. I sent them all to Vann T on Tuesday, I'll forward that email to you.
01-08-2019 03:48 AM
I think this might be similar to what I am seeing with mine as well.
The funny thing is that this only happens with Air Prisms. To make it worse this doesn't happen on one part of my nodes in a ring topology.
Let's say I have two nodes A and C with a B node in between them.
The odd thing is that sometimes both Air Prism backhauls on A and C will drop from the network and no longer able to access either of the Air Prisms that look to be connected to a cisco switch on those nodes, even if the air prism is set as the main PtP AP which it would still be possible to access if all of node B had went down but this wasn't the case.
To make matters even more confusing is that the switches' interface status for that port would show it's up but no data passing on top of no mac address in the table this would show the same on Node C as well.
01-08-2019 08:07 AM
Yeah, tough for a L2 switch to pass traffic without a MAC entry. We've got around 700 nodes on our LAN, these Ubiquiti devices are the only problematic ones we have, and only the ones at the last-last mile bridge. Again, the common denominator is a bridge to a switch, to another bridge to a switch. We can't be the only persons having this issue...
01-08-2019 08:44 AM
That's ironic, because we tried (3) different switches, and all 3 switches resulted in the same symptom. Those three switches are a Cisco 2960X, Ubiquiti US-8-150, and a (wait for it....) NETGEAR unmanaged switch.
01-08-2019 11:02 AM
No, I'm not 100% positive. But, STP is disabled on the radio, and there are no logs of a STP event on our switch(es). Additionally, the port interface that is the downlink to the radio is an access port, not a trunk port.
Is there something you have advice for that I should be doubler-checking in logs or configurationally? Thank you for taking the time to reply to my issue.
01-09-2019 07:16 AM
01-12-2019 07:38 AM
Network Mode is Bridge, configuration Mode is SImple. IP is static, it's a /24 network. MTU 1500. STP, Management VLAN, Auto IP Aliasing, DHCP Option 82 and IPv6 are all off.