v1.4.6 Bug reports and Comments

[bduley]

SNMP uptime show 263 days on firmware 1.4.6.
Firmware 1.4.5 does not have the problem.

snmpwalk -v 2c -c public 192.168.1.20

SNMPv2-MIB::sysDescr.0 = STRING: Netonix WS-6-MINI
SNMPv2-MIB::sysObjectID.0 = OID: SNMPv2-SMI::enterprises.46242
DISMAN-EVENT-MIB::sysUpTimeInstance = Timeticks: (2274583118) 263 days, 6:17:11.18

[Eric Stern]

I checked this on a few different switches, they appear to be fine. I tried a few things that could possibly cause this problem but was unable to duplicate it.

If you can describe a way to duplicate the problem I can look into it further.

[bearm]

I have a WS-24-400A that flaps the link on one of my APs.
Disabling Spanning Tree Protocol on that port is the only thing that resolved the issue.

I upgraded from 1.4.5rc2 to 1.4.6 in response, but firmware did not fix the problem.

Some days later I noticed repeating log messages on the port the affect AP is connected to.

Code: Select all

Jan 12 08:13:32 STP: set port 8 to discarding
Jan 12 08:13:32 STP: set port 8 to learning
Jan 12 08:13:32 STP: set port 8 to forwarding
Jan 12 08:24:17 STP: set port 8 to discarding
Jan 12 08:24:17 STP: set port 8 to learning
Jan 12 08:24:17 STP: set port 8 to forwarding


I find the link flaps or has loss ... the PPPoE session drop resembles the time periods of the log messages above.

As soon as I enabled STP on the port again, the flapping/session timeout problems started back up (after leaving STP off and having a stable link for over one hour).

My inquiry on the WISPA mailing list resulted in the following response:

I have had to disable STP on most of my netonix switches. They will start
flapping ports even if there are no redundant paths.
I need to let Chris know about this, it's been a real PITA, causing SINGLE
links to drop without warning. ...snipped...

In this leg of the network, I have a ToughSwitch, NanoBridge, and MikroTik at a remote location that connects back to AP, Netonix switch, Cisco Catalyst switch, and Cisco Router. The MikroTik is the PPPoE server at the remote location. For some reason link issues result in the PPPoE session timing out.

I do know for a fact that the issue arose when we migrated that AP to the Netonix in December. And that disabling STP on the port the AP hooks to did resolve the problem.

Thanks for your help (Chris, Eric, and the community).

[sirhc]

Well all I can say is I have 20+ WISP Switches in service and they all have RSTP active as well as Loop Protection and I have absolutely no issue at all with STP???

Below are 2 of my my largest/heaviest loaded switches, no flapping routes?????

Now I do run a large routed network so one switch does not feed another switch directly as there is a router at every tower???

However in my offices I have 2 or more WISP Switches hooked together and I have no flapping ports in any of my offices???

At Netlinx we have (2) WS-24-400A and (1) WS-12-250A with no issue.

At RF Armor / Netonix we have a WS-12-250-AC and several other switches we use to test firmware and serve the office and no flapping ports?

Also if you have another switch in the mix like a ToughSWITCH which does have RSTP broken who knows???

Maybe try enabling RSTP on 1 port at a time until you find the issue? Might take days to discover the port or ports that lead you to the issue?

CLICK IMAGES BELOW TO VIEW FULL SIZE

[tma]

Chris will probably disagree with me but my experience with RSTP (which comes from Switches other than Netonix) tells me to have it active only on the ports where you need it (because they are part of a ring) and to turn it off entirely if you don't have such ports. I.e. don't turn it on only because somebody may plug a cable wrongly some day. Further, if you have redundant paths between your sites - and they aren't routed, so you need RSTP - make sure you have minimal packet loss and prevent saturation on your wireless links. This is because, if you loose some RSTP "Hellos", you will likely see flapping.

That aside, you mentioned that the port is connected to an "AP". If this is a UBNT AP which is (most likely) configured in bridge mode, you can turn (R)STP on and off on the AP bridge as well. That detail may make a difference, and if STP is set to ON, I'd suggest to turn STP off on the AP because it is enough for the AP to pass STP packets on - it doesn't have to be an active participant in the RSTP topology unless this AP is a junction point to redundant paths, which is, by definition, not the case on APs in a PTP link and not for PMTP APs with leaf-only clients.

Also, I've seen learning states happen on a port which were caused by link states changing on other ports, even at remote sites, which is a PITA. Again, this is not Netonix specific, but some other switches have more RSTP port configuration options to mitigate such effects. We have modified the UBNT Airmax ebtables rules to not pass RSTP packets between sites if they are in different STP domains. I.e., if you can, split your switched networks into smaller STP domains and prevent any STP communication between them.

[sirhc]

Chris will probably disagree with me

Actually most all your suggestions are good stuff Thomas and valued advice to many.

I simply have more things on than I really need (almost everything is turned ON) as my WISP is a test bed for Netonix.

But in exchange that guy Chris aka sirhc lets me call on him 24/7 if I need help.

[Ludvik]

Port 1 right LED indicator on WS-6-MINI still lights up - little confusing. This indicating 24V POE OUT - but on this port doesn't exist.

sorry for my english.

[lligetfa]

Port 1 right LED indicator on WS-6-MINI still lights up - little confusing. This indicating 24V POE OUT - but on this port doesn't exist.

sorry for my english.

No PoE out on port 1. It is a power indicator.

[bearm]

Well all I can say is I have 20+ WISP Switches in service and they all have RSTP active as well as Loop Protection and I have absolutely no issue at all with STP???

Below are 2 of my my largest/heaviest loaded switches, no flapping routes?????

Now I do run a large routed network so one switch does not feed another switch directly as there is a router at every tower???

However in my offices I have 2 or more WISP Switches hooked together and I have no flapping ports in any of my offices???

At Netlinx we have (2) WS-24-400A and (1) WS-12-250A with no issue.

At RF Armor / Netonix we have a WS-12-250-AC and several other switches we use to test firmware and serve the office and no flapping ports?

Also if you have another switch in the mix like a ToughSWITCH which does have RSTP broken who knows???

Maybe try enabling RSTP on 1 port at a time until you find the issue? Might take days to discover the port or ports that lead you to the issue?
...snipped...

While my radios/links are in bridge mode, they patch into a switch which has the links segregated into VLANs on the switches. So it's not one huge Layer2 switching mess of a network.

The Ubnt ToughSwitch could be the cause.

I've isolated/resolved the issue by disabling STP on one port, but while I mention the Netonix side that's because it's configurable and provides helpful logging whereas the ToughSwitch does not.
On the ToughSwitches, STP state information is a blackhole (as compared to Cisco IOS and Netonix). I have local and remote logging enabled on ToughSwitch and get squat for information.

(At least right now) I literally have 1 problem (the 2.4GHz AP/link) out of many backhaul links that terminate/are fed via the WS-24-400A at the main office. Not bad!

You're not seeing issues with Netonix-to-Netonix STP.
Do you have any test beds where ToughSwitches are interop tested?

At this point I'm stating what solved my issues and when it began, not placing blame on Netonix. I don't know definitively which item, configuration, device is at fault yet.

[sirhc]

Do you have any test beds where ToughSwitches are interop tested?

I do not have any ToughSwitches in my network, sorry.

Here is one post about TS RSTP issue:
viewtopic.php?f=17&t=570&p=4066&hilit=+rstp+broken#p4066