Ok, full top down view first. Because this has been a disaster and nightmare for years.
From the perspective of the TA5k, the network is fairly simple. Core router which is connected to multiple peers, core switch which is connected to the core router, Adtran connected to the switch with 2 10Gbe ports. I'll go over cards and vendors in the description.
The issue is, if any of you use the adtran platform, that in adtran you first need to lay out their queues and priority mapping. By default, they have 4 cos queues, but support 8. We have them set to 8. And have rebooted the switch modules for 8 to be effective. Now, here is the actual problem I am trying to work with them on. For GPON customers, we would like to shape on queue 0, which is mapped to vlan priority 0. This makes sense for data (internet) traffic. But when the shaper is set on queue 0, customers get gigabit service. I worked around this issue by shaping on multiple queues in the single shapers (0-2), but this needs to be cleaned up on our network. We shouldn't be using 3 queues just for standard data. On the phone with support, we started looking at what adtran saw on the switch module. This was in 2013. They told me that my "core switch must be sending vlan priority set to 1 on these data vlans because they only see the traffic at priority 1". In 2013, i decided "ok, uh... maybe my switch is jank, but ok".
Now it is 2018. I have replaced our core router a few times, but it doesn't actually matter in this scenario. The core switch got replaced to an HP 5406R (i know this is now "aruba" but it was still HP when we did it..... I also know this isnt a "core switch" and more of an enterprise switch, but i needed PoE+ and we have a Juniper MX480 for use with the stuff that won't be on the HP switch, it just didnt have the cards in it yet. So lets just leave that for now.)
After 5 years, 2 core router installs/upgrades, and a new switch which is connected to the adtran, same issue. Cannot use shapers on queue 0. Got on the phone with tech support, troubleshooting for 5 days straight, they are still saying "that switch is setting the priority to 1".
So, got my laptop onto the switch. Tagged a few vlans (the same vlans used on the adtran internet data portion), ran tcpdump on the physical interface (to see the vlan tags) and started changing the priorities. At this point, the only difference in my laptop and the adtran as far as the connection to the switch, router, and internet is that the laptop has a 1gig port only. Every time i change the vlan priority on the vlan in the hp/aruba, i immediately see that vlan priority on the dump. I mean, no shit. But i wanted to verify this. And i saw what i expected to see. The switch is not doing anything weird.
Back on the phone with tech support from adtran, and they cannot see me changing (overriding) the vlan priority... from 0 to 3. This part is important. If i set the vlan priority in the aruba to 0, 1, 2, or 3, they see vlan priority 1. After 3 they see the correct value.
Next, I humored them by switching the active 10gig data port to another 10gig port on the switch. Same issue. Then i switched the sfp+s on both ends. Same issue. I would say the only thing we haven't changed is the switch module on the adtran, but that isn't true because between 2013 and now, we have upgraded our adtran shelves from SM25s (the 2 port 1gig sfp and 2 port 10gig xfp) to SM40s (4 port 10gig sfp+). So we have changed the cards. Literally, at this point, i dont know of anything that hasn't changed. We have physically moved our equipment from one building to another. lol.
They are still pointing fingers at my network gear. For now, I am allowing it, because I have the Juniper that we are going to cut the core router and part of the switching over to. The only reason I haven't done it yet is because it weighs approximately 7 solar masses and just haven't been looking forward to breaking my back getting it racked up. But I told adtran i will do that. So... now the real issue.
I have no doubt in my mind whatsoever at all... that when swing these links over to the juniper, that we will have the same issue. What the hell will be my next step? All other equipment (adtran is not our only vendor) works as expected, plugged in to the same switch. I don't actually blame adtran for pointing their fingers at my gear. Especially when we have changed the switch modules on the adtran and still have the same issue. If I were adtran, i would probably say "well, thats 2 different models of switch modules doing the same thing, so it isnt us" or something. But what is next? A 10gig tap? I seriously doubt my company is going to boner up the money to buy a 10gig tap when we all feel like this issue is on adtran.
I will not tl;dr. Just don't read if it is too long.
No comments:
Post a Comment