WS-12-DC stops passing traffic when B11 link goes up

[brock]

I think this may be related to some issues reported in this thread:

viewtopic.php?f=17&t=1895&hilit=v1.4.2+bug+reports&start=20

We had a bad problem yesterday when we tried to install a new B11 link on a WS-12-DC running 1.4.2. It has been working fine on 1.4.2 so we hadn't upgraded it. We are planning to upgrade today but I want to better understand what could have been happening.

When the B11 was powered on, things were fine. As soon as it linked wirelessly to the other B11, we lost all communication with the router attached to the switch. Everything was going through one port (router on a stick) via VLANs and untagged on the B11 interface (port 3, 48HV).

When we set up an untagged VLAN between ports 8 (connected to another router interface) and port 3 (connected to the B11 link), we could then get into the router ONLY via the B11 link. Attempts to ping the switch, the access points, the other backhauls, etc, from the router at that point all failed until the B11 link was taken down.

This even happened when the B11 was connected to an untagged VLAN on a port where the VLAN was excluded on ALL other ports (i.e. the B11 was talking to the switch but not the router or anything else). Because of this we have to conclude that the router was not the issue.

Now, in the thread I linked above, supposedly going to 1.4.3rcX fixed something very similar for another user. However, this seems like it's a switch issue, and Chris has said in other threads that the Netonix firmware does not affect switching. Is it known what the problem actually was in that thread above and how it was fixed? There's nothing in the 1.4.3rcX release notes. We spent hours troubleshooting this yesterday on our high sites with nothing but customer service disruption to show for it so before trying again I'd really like to understand what happened.

[sirhc]

I did say the CPU has no effect on switching capacity and I have said our firmware just configures and controls the switch core and does not handle the packet handling, you are correct on this so far.

HOWEVER the bugs that are in v1.4.1 through v1.4.3 can affect the CPU which intern can affect the CPU's ability to control the switch as its control Daemons are not getting enough CPU time.

Plus you want v1.4.4 at LEAST as it has some new security patches but v1.4.5rc2 simply has bug fixes reported in v1.4.4 so you should use v1.4.5rc2

Please upgrade to v1.4.5rc2 and report if the issue is still there.

[brock]

We will try 1.4.5rc2. We are also bringing a spare WS-12-DC and will be trying downgrading the firmware on the B11 from 1.4.1 to 1.3.1 if nothing else seems to fix it.

[brock]

Today we tested from the base of the tower to a laptop (not to the other end of the link) and there appeared to be no problem, even before upgrading firmware. So it seems to be something at the other tower site (Mikrotik router or ToughSwitch) that's causing the problem. It's linked to a different backhaul location via a B5c right now (and has been for months) without issue. However, that is going straight into a Mikrotik, no external switch (Netonix, UBNT, or otherwise). Doesn't seem to be the B11.

We upgraded to 1.4.5rc2 also. Will test more next week.