VLAN issues with Intel I350 GigE NICs

[Newfoundland.Republic]

My X9DR3-LN4F+ uses the Intel I350 GigE NICs. There was a bug on VLAN problems reported for FreeBSD (and also pfSense).

Was this fixed in FreeNAS 11.2 U7?

 

[jlw52761]

I also use two of the Intel I350 Quad Port cards, and am experiencing an issue with VLAN trunking and these cards. I am on FreeNAS 11.3 U1 and can confirm the situation still exists.
I'm actually writing a post on this in the forums hoping to get some traction on this as not being able to use VLAN trunking is horrible.

 

[Elliot Dierksen]

I am running 11.2-U7 and doing VLAN's in an LACP LAGG on i350's without any issues. My units are bare metal, and storage only. No virtual machines.

 

[y4z]

I am running FreeNAS-11.3-U2.1, and I too am having issues with vlan interfaces (I350).
Oddly enough I was not having issues until I added a third VLANinterface.

Now I am getting a flood of TCP RESET ACKs whenever traffic is tagged. Less than ideal to say the least ;)

 

[Newfoundland.Republic]

I ended up deleting all the interfaces when I went from 11.2 U7 to 11.3 and everything seems to have cleared up. I rebooted when I did this to make sure everything was stable.

 

[y4z]

I ended up deleting all the interfaces when I went from 11.2 U7 to 11.3 and everything seems to have cleared up. I rebooted when I did this to make sure everything was stable.

Thanks for the followup. I'll try this over the weekend and report back.

 

[y4z]

Still having issues. I deleted all of my network configurations through the TUI (/etc/netcli), then powered off the system.
Afterwards, I replaced my second HBA with an Intel PRO/1000PT DP NIC. I then configured management on a newly created lagg0 (lacp, em0+em1), and vlan109 for data (lagg1, lacp, igb0+igb1).

Seeing TCP RESET ACKs when traffic goes through the default route of the FreeNAS box (regardless of interface). Therefore, my issue is unrelated.

 

[Elliot Dierksen]

Have you checked your switch config to make sure it and FreeNAS agree on what vlan should be untaggged? This sounds like a config mismatch.

 

[y4z]

Have you checked your switch config to make sure it and FreeNAS agree on what vlan should be untaggged? This sounds like a config mismatch.

Unfortunately I have. My first thought was that there was something wrong with my firewall (i.e router, pfSense box). However, I do not experience disconnects when testing Debian Buster live media on the FreeNAS box. Additionally, I tried running new cables and changing switches (Cisco 2960L to a TP-Link T2600G).

Here are my current switchport configs for the T2600G:

Code:

sw2#sh run int gi 1/0/3
!T2600G-28TS
interface 1/0/3
  description "fs1_mgmt"
  switchport general allowed vlan 10 untagged
  switchport pvid 10
  no switchport general allowed vlan 1

  channel-group 1 mode active
#
sw2#sh run int gi 1/0/4
!T2600G-28TS
interface 1/0/4
  description "fs1_mgmt"
  switchport general allowed vlan 10 untagged
  switchport pvid 10
  no switchport general allowed vlan 1

  channel-group 1 mode active
#
sw2#sh run int gi 1/0/7
!T2600G-28TS
interface 1/0/7
  description "fs1_file"
  switchport general allowed vlan 10-21 tagged
  switchport acceptable frame tagged
  switchport pvid 666
  no switchport general allowed vlan 1

  channel-group 2 mode active
#
sw2#sh run int gi 1/0/8
!T2600G-28TS
interface 1/0/8
  description "fs1_file"
  switchport general allowed vlan 10-21 tagged
  switchport acceptable frame tagged
  switchport pvid 666
  no switchport general allowed vlan 1

  channel-group 2 mode active
sw2#sh run
...
#
interface port-channel 1
  description "fs1_mgmt"
  switchport general allowed vlan 10 untagged
  switchport pvid 10
  no switchport general allowed vlan 1
#
interface port-channel 2
  description "fs1_file"
  switchport general allowed vlan 10-21 tagged
  switchport acceptable frame tagged
  switchport pvid 666
  no switchport general allowed vlan 1
...

Where VLAN 666 is my blackhole, and VLAN 10 is management.

Last night I tried clearing my link aggregations a second time from the /etc/netcli menu and experienced a kernel panic. The IMM on the box is not reporting hardware issues... Might be dealing with ghosts

 

[Elliot Dierksen]

Looking at port-channel 2, it expects all tagged traffic. That would mean you would not want to use the LAGG interface itself, but only VLAN sub-interfaces. I assume that is the LAGG that you are using for the additional VLAN's. Is that accurate?

 

[y4z]

That is correct, I have created vlanX interfaces off of that LAGG interface. I do not have any IP configuration on the LAGG interface itself.

 

[Elliot Dierksen]

Do you need VLAN 10 on that port-channel 2? I wonder if that is messing up the mac address table either on the switch or in FreeNAS. Try removing vlan 10 and see if that helps. If not, you may have to make that into a single port channel (you could use 4 members). I don't know that for sure, but that seems suspicious to me.

 

[y4z]

Quick update: I eventually caved in and installed Debian. At this point I am fairly certain my issue was a bug
Thank you Elliot, appreciate your efforts trying to troubleshoot this with me.