MikroTik

Hi,

I have been having an issue with 2 CRS running RouterOS where I have 9 UBIQ Prism APs on Each. I am seeing random reboots on the UBIQs but not on any access ports or PTPs. I have a site with a UBIQ GPON and it does not reboot. The Reboots appear to be soft reboots and i believe caused by a Storm. Running out of ideas and glad to post Configs if anyone thinks they can help.

Respectfully,

Eddie

Hi Eddie,

might be able to help. I've used Mikrotik and Ubiquiti together a lot. Configurations would be nice but definitely check for PoE capacity of the switches and if it can handle all of your connected devices. With that said please let me know:

1) If the Prism APs (which model specifically?) are powered from the switch or using their injector.

2) How do you come to the conclusion that those are soft reboots. Are you basing yourself on the ubiquiti error message that says unexpected reboot? Basically a lack thereof.

3) Are ALL prismd affected?

4) Are any other PoE devices affected on the switch? Based on what you're saying I think that only the prisms are affected but I wanted to clarify

Best

Thank you. All UBIQ APs are on each own PoE. Currently @ 1 site I have a CRS328 being Feed from a Brocade MLX via a Licensed Link and the CRS is feed a Brocade CES2048 via a UBIQ 24Ghz PTP. When i leave the current config (made a few changes now) i will have random reboots @ for most Prism Gen2 5Ghz running 8.7.11 on the CRS326 and the Brocade CES via the 24Ghz. I will have 1 UBIQ ISO @ the Office off the Brocade MLX that will also reboot randomly. I came to the soft reboots because i will run a ping on a AP and see that when it drops it will only lose 4-6 pings and i will look at uptime and it just reset. Doubt its a full reboot but who knows. Now, if I remove the CRS326 and reconnect the Cisco 3750X then all of this goes away. None of the bases will reboot be it soft or a full reboot. I have left the Cisco up for 2 weeks and no reboots.

Ok, can I have a clarification regarding the topology. So far I understand the following:

Brocade MLX -> PTP Link (licensed) -> CRS 328 (or 326? you mention both 328 and 326) -> 24ghz PTP -> Brocade CES2048

I see that reboots also happen on the Brocade devices so it's probably nothing to do with the Mikrotik but configuration-wise I'm noticing many vlan interfaces whose purpose I don't really understand. Seems like you're combining old and new vlan configuration (I'm not familiar with the old way of vlan configuration, I'm saying that based on what I've read here and there, so please correct me if I'm wrong).

Are you sure that the uptime timer resets on the ubiquiti devices? Because 4-6 missed pings do not add up to a reboot, not even a soft one. I tested it on a LBE AC Gen 2 running 8.7.11 and it dropped 14 packets (6 * Request timed out, 8 * Destination host unreachable) but I can only imagine that a reboot on a Rocket AirPrism will take longer (I don't want to reboot one of our Rockets so as not to cause downtime to the clients). Have you noticed any random auto-negotiations on the interfaces? Can you post log files for the relevant interfaces?

When you replace the crs328 with the Cisco switch, the problem goes away. This begs the question: Are you using the same cables to connect the Ubiquiti devices? I presume that the PoE cable
is the same, but what about the cables running from the LAN port of the injector to the switch? Anyway a reboot that owes itself to configuration (especially when drawing power from the injector and not directly from the switch) could be cause by anything overencumbering the device CPU (e.g., we've had reboots on cambium radios due to cpu usage spikes caused by the frequency analyzer we had forgotten to turn off ). Could there be a layer 2 storm occurring in the vlan the airprisms are in (maybe an oversight in the vlan configuration of mikrotik which is not present in the cisco config)? This would require a more detailed topology and configurations from the rest of the devices. You can also try to enable graphs in the crs328 where you can notice any abnormal traffic patterns (https://help.mikrotik.com/docs/display/ROS/Graphing).

Lastly, what firmware is running on Mikrotik? I see that you're using development channel for updates which might not be ideal for production environments. Maybe install firmware from the stable channel?

Hi,

Sorry its a a CRS328. Minus the 2 Fiber connections on both Cisco and MikroTik its Port for Port (Straight swap from 1 to 1 via ethernet). Currently running the last Stable Version on version 6. Yes, the uptime on the Prism APs resets to 1min or 1 hr depending on reboot. Brocade MLX -> PTP Link (licensed) -> CRS 328 -> 24ghz PTP -> Brocade CES2048 (Correct). This only occurs when the CRS328 replaces the Cisco3750. All APs need to be in trunk mode with VLANs 33,35,116. If its a Config issue its my config in the CRS because its not an issue with my Cisco Config. Same on the CRS354 as when i moves cables back to the Cisco 3750X there all issues go away. I believe the issue to be a Storm and its due to VLAN 1 (VLAN Filtering PVID=1) I have changed the "Admit All to only tagged" and so far 16/18 bases have stayed up. Its clear thats it my config within the CRS and not networking as again when on Cisco's i have no issues at all. Hope that Helps. Thanks for helping.

Is there redundancy in the network? I see stp is disabled on the crs328. And PLEASE check if any of the radios has stp enabled and disable it if you don't need it. Maybe there are topology changes in the network (e.g., a client antenna with stp enabled comes online) and the radios need to reconverge. I don't know how AirPrisms would handle reconvergence because we don't have stp enabled on client APs but based on what I'm reading onllne a reboot because of stp reconvergence is possible. If stp is indeed enabled on the ubiquiti devices you can check which bridge is the root and if that changes when you notice reboots.

If spanning-tree is the problem, then that explains why cisco does not have it. If cisco is running the default rpvst (and not mst) it will not pass on normal rstp bpdus (I think that is the case) so the radios will never be able to send bpdus to the other radios so no reconvergence => no reboots.

Makes sense. All APs have STP off and all clients aswell (Default but you never know). VLANs 33,35,55,116 all have RSTP on them on both Brocade MLX and Brocade CES ( MLX > PTP License - crs328 > PTP 24Ghz - CES. I have tried turning on RSTP on the Bridge but didnt help and cant say i programmed anything correctly on it though. The issue i see and hope you can help is that when you Enable VLAN Filtering the PVID = 1 (admit all) and same on all the Bridge Ports that are Trunks. I have nothing on either Brocades on VLAN 1. So this is where i cant wrap my head around it. Ideas?

I don't think VLAN 1 is related at all. As a matter of fact setting frame-types=admit-only-vlan-tagged, as per the wiki (https://help.mikrotik.com/docs/display/ ... +Switching) effectively disables default VLAN 1. Anyway the PVID value is used on access and hybrid ports and instructs the switch to place untagged traffic ingressing that port to the specified vlan so that it traverses the uplink trunk tagged with that particular vlan-id. Having frame-types=admit-only-vlan-tagged stops any untagged traffic except from special traffic that is intended for the CPU (e.g., BPDUs which are sent untagged but are not dropped by the admit only vlan tagged rule). Have you checked the ubiquiti logs (it's a little symbol on the bottom left of the UI for Rocket AC)? Anything interesting over there?

What we know so far:
1) Ubiquiti APs reboot randomly
2) Cabling and PoE injectors not an issue
3) Mikrotik hardware most likely not problematic since it is not the power source for the devices
4) Most likely nothing to do with RF environment since changing the switch resolves the issue
5) Most likely a configuration thing

I will take up on the 5th point and ask anybody on the forum what configuration problem could cause reboots on a device, I'm curious myself. It should be something that causes spikes on CPU usage (like the issue with the cambium spectrum analyzer I mentioned previously) but I don't know what else it could be other than stp reconvergence hogging on the cpu.

So, I'd suggest taking a look at the logs (both on mikrotik and ubiquiti) and post anything interesting over here. Also, as a side note, I think you should clear up your vlan configuration. It seems to me that all those vlan interfaces (except the management interface) have no purpose other than make the configuration more complicated and difficult to read. Couldn't that be done with tagging and untagging vlans with the bridge vlan function (the wiki is again really illuminating https://help.mikrotik.com/docs/display/ ... p+features).