I have 3 routers:
- RB3011 (main router, internet gateway)
- RB941-2nD (remote1)
- RB2011UiAS-2HnD (remote2)

Both remote routers are connected to main router via double long Ethernet cables, and so for redundancy and performance reasons I'm using bonding with "balance rr" mode.

After upgrade 6.37.3 -> 6.38.1 on all three routers I could not manage to restore TCP/IP traffic between routers, the communication just stopped. I could see activity at bonding interfaces, each individual router worked fine when I connected directly with a cable - just there was no TCP communication between routers at all.

After downgrade to 6.37.4 - everything went back to normal.

Any idea what could be wrong ??

Sorry to bump - and add details: all routers have bonding interfaces attached to local bridges. Is bridge routing broken in 6.38.1? Or this is a wrong place to ask?

I'm fighting a similar problem with an hAP ac that I just set up and installed for a customer. The configuration is dead simple, mostly stock configuration plus an L2TP/IPsec VPN gateway with RADIUS for auth... It did have bonded interfaces for a server, but even with that removed, I'm seeing a bunch of broken packets that appear to be getting damaged while traversing the bridge (packet length errors). The problem did not occur over the switch chip when everything was wired while testing, but it's unusable over wireless or through the VPN (the only methods the customer uses), which requires bridging.

Did a downgrade help by any chance? I'm afraid to do it remotely with the severe packet loss on mine...

Yes, I downgraded all routers to 6.37.4 and everything works fine, without any changes in configuration at all. So clearly there something seriously wrong (or different) in 6.38.1

Did you read the release notes? Noticed anything about spanning tree? Did you try to disable spanning tree on the bridge?

Did you read the release notes? Noticed anything about spanning tree? Did you try to disable spanning tree on the bridge?

Well, that's not clear. According to Wiki the spanning tree protocol applies to CSR models, which is not my case. Are you saying every router would get that change?

Funny thing, I had this same problem with a cAP2N. After upgrade cannot access it from the wired side, It showed connected Ethernet, and it even passed traffic for two other Virtual AP's onto VLANs. But I could not access it from the wired side on any method.
I connected to the AP through wireless, and I could connect with the "tikAPP" on the android phone. Could change settings etc. Ironically my phone got an IP address from a distant DHCP server that was connected in such... cAP2n >wired< RB493 >wired< CRS125 The CRS 125 did the DHCP.

I could not access anything other than this cAP while being connected. While sitting reading your post I thought, I'll try this and I changed the Bridge on the 493 that included the Ethernet port 2 that connected to the cAP from rstp to NONE and then all of the sudden I could access the cAP2n with no problems. The Bridge on the cAP2n is set to rstp still. Switching back the bridge on the 493 to rstp from none breaks connectivity.

fyi - power cycle on the cAP2n prior to any of this did not fix the issue, nor did a reboot when connected from the wireless side.

Any ideas as to why it was working fine before the upgrade and then not afterwards?

Just a side note, rstp is enabled on the cAP2n on 3 bridges, my private, and two vlan bridges that connected to virtual AP's. Just disabling on the 493 solved my problem.