Problem connecting 2 switches after VLAN change, and link agregation for Hyper-V

[dehcbad25]

First, let me tell you that I messed up, and I am getting probably a few things confused, so most likely it is something silly that I am doing.

I have 4 v1910 switches, 2x48G, 1x24G amd 1X16G.

The 2X48G sit in the cable rack, and I had them connected to each other (if I remember correctly in stack mode, but that might not be true...it has been a long time). Lets call them switch 1 and switch 2.

The 24G is in the server rack, and connects to switch 1 in port 1.

The 16G is in my office and connects in another port.

I had 3VLANs, but only switch 1 and switch 2 used them (V1 which was default, and a Guest VLAN(2)), The 24G also had a 3rd VLAN for storage only.

I had to switch the subnet in the network and since I had to do sections at a time I created another VLAN (224).

When I setup all switches I had switches running so I did my setup relaxed. When I switched the subnet I had a ton of other work too, and I did the whole thing in 2 days, so I was shooting for the fastest resolution.

Ever since then, switch 1 and switch 2 do not talk. Since I don't need all other VLAN at the moment (I have other ways to isolate guests since we have Meraki router and AP), I tagged all ports as VLAN 224. Again, this works from the server swith to switch 1, and from the switch in my office to switch 1. But switch 2 does not connect to 1.

I am thinking that the problem is untagged traffic, or something along those lines, but I would like to understand more clearly what I am doing wrong. Since I have the network running, I don't get a lot of time anymore to just "try" since I don't want to cause downtime.

My other issue, is that when I was configuring the VLAN I selected all ports in the server switch and I forgot I had a link agregation with LACP. I have a Hyper-V server, and I am using NIC Teaming. I had this working before, but unlessI set it as independent it won't work now, and that just has me a little confused.

My next step is to remove VLAN 224 and go back to default 1, but not sure if it will fix the problem, and I will have to go back to the office do this work at 2 AM....and walk to the office the next day looking relaxed (which is not happening, I like to sleep)

Again, I must admit, that even though I know VLAN, and how they work theoretically, I don't have much experience using them besides simpler setups, so some help with this will be greatly appreciated.

Also, I used an article (I think from HP) to setup Link Agregation and LACP but I was not able to find it

[dehcbad25]

OK, I did not get to the point of reseting the switch, but it seems it would have solved the problem.

When I did the initial configuration 3 years ago, I tried doing link agreagation between the 2 switches to have 2 Gbps connection, but whenever I did it would create a loop instead. I removed the link agregation, but apparently it came back.

I did have that problem with the config before. Some settings seem to revert back after a reboot. Maybe I am forgetting to save the configuration, maybe there is a problem in the memory. No idea.

The bad config was in switch 1.

Switch 1, port 48 Trunk, Tagged membership 224, PVID 1

Switch 2, port 47 Trunk, Tagged membership 224, PVID 1

I think the reason I did not cought the problem earlier was that I was concentrating in switch 2, and I don't have many free ports in switch 1, so I did not try another port.

Now lets see if I can resolve the problem with link agreagation

[parnassus]

---

dehcbad25 wrote: When I did the initial configuration 3 years ago, I tried doing link agreagation between the 2 switches to have 2 Gbps connection, but whenever I did it would create a loop instead. I removed the link agregation, but apparently it came back.

That's quite strange...deploying a LAG (Port Trunking) with only two dedicated ports per Switch between two HPE OfficeConnect 1910-48G should be very easy, a 10 minutes GUI setup task if you plan it correctly (provided that you know what to do and when to do it...and that you start with member ports in a known default state, not messed up, as example: physically connecting both ends with necessary copper/optical cables is the very last step to perform...before that you should correctly setup LAGs by adding Member ports and deciding Link Aggregation Algorithm Static=No Protocol/Dynamic=LACP, set LAG Port Type of type "Trunk" and eventually permit VLANs, other than VLAN 1, you need to flow through the created LAG).

Which software version your switches are currently running on? that's a thing to take care of too...

[Vince-Whirlwind]

? You created a new subnet on a new VLAN and added it to both switches and then "switch 1 and switch 2 do not talk"? What does that mean? - you should detail precisely which tests fail - physical connectivity? LLDP? ICMP? switch to switch or device to device within a subnet using a path through both switches, device to device across subnets within a switch, device to device across subnets across switches, etc.... and the context of switchport configuration, VLAN interface configuration, and device IP configuration.

Document your network: switches, switchport IDs, IP addressing, and VLAN assignment. It's far less effort than trying to describe it in words. And *much* more accurate.

[dehcbad25]

Yeah, ideally I would document a lot of stuff, but sometimes the problem of being 1-man show is that you run out of time to do things properly

For the last 2 years however I have been using OneNote and at the very minimum copying and pasting from web GUI (not from the swich since I rarely had to go in it).

However, with all honestely, most of the changes i doubt would ever get properly documented all the way. I know I sing a familiar tune, but there really is no time, and after all I am the only person working on it

[dehcbad25]

Both switches were updated a month ago to the latest version. At the very minimum they are runnning the same version.

Last night I tried again and I had the same problem, although this time it was better, the switch automatically blocked the ports so they did not loop. What I did was:

Go in switch 2, setup port 45 and 47 as part of a Link Agregation Group with Dynamic (LACP). Setup both ports as trunk

Go into switch 1, setup port 46 and 48 as a Link Agregation GRoup with Dynamic (LACP). Setup both ports as trunk

switch2 port 47 was connected already to switch 1 port 48. When I connected switch 2 port 45 to switch 1 port 46 I lost connection to switch 2.

By this time it was already late so I just disconnected 45 to 46 and changed the port back to access. (I was doing this after 4 AM last night)

[dehcbad25]

Big update. I have been having problems with our AP forwarding DHCP request in the network. I had fixed all issues I had previously by creating proper trunk ports and assigning the trunk ports to PVID 1, but the AP still had problems (not all the time, not always either, so it was confusing)

Finally last night I came back to the office after 11 PM, and I decided to initialize the configuration of switch 2. With swithc 2 back to default, I configured again the trunk port to switch 1, left all ports as default, and then I proceeded to switch the other 3 switches all to PVID 1, and remove any complex setup.

I removed the extra VLAN interface as well (since we don't neeed anymore the first subnet). and I made our 192.168.224.0/24 subnet run in VLAN1 instead of 224.

It took a whilte because I had to physically keep chaning connected ports to update them (I use a Surface Pro, so no serial for console, and I can't find my USB to serial adapter), but in the end I simplified the configuration. After I verified it all worked I saved configuration, restarted the switches (one of the switch had a problem on restart) but everything works (except the AP still have problems with the DHCP request)

I was able to setup LACP with the Hyper-V port again, and it was easy and worked. So I think there was something weird in the config from before.

Only thing left is that I could not setup Link Agregation between the switches, but I am thinking on doing that after I find fiber modules and do it directly at the fiber port.

Any recommendation for the fiber module?

As far as the DHCP request go, I am beginning to suspect the culprit is out Meraki to non Meraki VPN. Somehow the request might be trying to go over the VPN link. I have very little information about what is in the other link, and it was totally not my idea to switch the LAN IP address either to link VPN. I have a feeling there is a DHCP server in the other end. I will test it tomorrow by removing the VPN link and checking if the issue persist.

[Vince-Whirlwind]

I don't remember using 1910s, but is "dynamic" the only option?

I recall on some switches, the Dynamic option would not initiate a LACP negotiation - one side had to be set to a different option.

[dehcbad25]

I figured out the problem with DHCP. It was actually that I was out of DHCP available IP addresses. The network change broke a lot of the reservations so I had servers getting regular IP leases instead of the reserved ones.

Some I had fixed during the change, but because lease time was 3 days it was exhausted. I switched to 4 hours, once leases expired then I started cleaning what should have not be getting IP (the guest network had to run in isolation and somehow it was connected to the regular for a while). Now I am down to 37% usage again (display statistics in DHCP...I will not forget to check that again)

So, totally not related to the switch, but I do think it might have been caused by the VLAN problem I had.

I will work on the link agregation between switches next. I just got a new batch of thin cables for the server room :)