Configuring VLANS Using PowerConnect 5448 & Hyper-V 2012 R2 vNICs

Question

Setting up new R710 servers to upgrade existing 2008R2 Hyper-V to a 2012R2 Hyper-V and current network was inherited and is very basic and all physical using default layer 2 switch VLAN (1), no vSwitch in Hyper-V just a single PowerConnect 5448 .

I am having trouble getting VLAN connectivity in the upgrade even tested right from the switch to the Hyper-v vNICs once I enabled VLANS in vNICs in Hyper-V. My objective is to create a few vNICs on each host that are utilizing a large team of 6 pNICs (mgmt., backup, live migration). Teamed using native 2012R2 with NO VLANs on the Team. Created a vSwitch then created vmnetworkadapter in Powershell and assign tagged VLANs to the management vNICs accordingly, while the VM traffic uses the native (default VLAN1) then on the Powerconnect 5448 we are using LACP for the Teams and I trunked the associated LAG channel #3. I did not do anything on the individual ports that make up the LAG believing the LAG VLAN config superseded any of that. Do not pay attention to DMZ Team, it is future once this gets working to allow a few limited VM's to run in the DMZ, while also having the ability to live migrate, backup etc. DMZ is a physically separate switch, hence the separate team and dedicated pair of pNICs.

Here is my configuration, please advise.

Hostmanagement IP - 192.168.2.9/24

LiveMigration IP - 172.16.16.5/24

Backup IP - TBD (Determine once I figure out this problem)

Upgrade Test VM - 192.168.2.131/24 (Using untagged native VLAN1)

Dell PowerConnect Switch Mgmt Interface (VLAN1) - 192.168.2.254

Upgrade server #2 used for testing on g29 switch port - 192.168.2.71

CH3 is made up of 6 ports (6,7,9,12,13,14), g29 is a test server in the mix. Show VLAN has CH3 and G29 both having VLAN 1 & 20.

From the powerconnect switch I can't ping the hostmanagement IP 2.9 and this 2.9 host has no connectivity at all from it or to it. I know there is some IP overlapping, but with VLANs it shouldn't be an issue. Ideally I plan on fixing this, but I am remote to the servers and having a hard enough time with this one using the iDRAC. Haven't tried the LiveMigration vNIC yet bc I couldn't get management to work. 2.9 worked fine until I applied the VLAN in Hyper-V via powershell.

2.9 can't ping the VM 2.131 which I would expect since they are on different VLANS.

I believe the configurations are all correct, but my testing is flawed, please enlighten me.

Also if there is an easier way for me to test this new implementation, please advise.

Shawn16 · Answer

Pics didn't post of configs, reposting them.

DELL-Josh Cr · Answer

Hi, Do the VMs communicate with each other? Can the switch communicate with them if they are not teamed? Can the host ping the switch?

Shawn16 · Answer

Only have 1 test vm on the new host and as stated its on the native VLAN (1) and can communicate to all other vm's in production and test. Not quite understanding the team question, native VLAN is running over the team so doesn't appear to be an issue. Host management is VLAN 20, switch is on the native VLAN (1). I can't change that remotely or I will get locked out.

The issue is around the defined VLANs, not the native.

DELL-Josh Cr · Answer

If you do a tracert where do the pings stop?

Shawn16 · Answer

I might have rectified this issue. I suspected the IP scheme was creating some confusion so I did the following and established a unique subnet:

Created 2nd management IP on PowerConnect switch on VLAN20 (For Testing) - 10.10.10.1/24

On HostManagment vNIC on Hyper-V host gave IP - 10.10.10.2/24

Both can ping each other fine now. I just have to get the 2nd upgrade server back online, got locked out after I created a vNIC and it wiped out the pNIC IP :-(. That will validate host to host comms. on the HostManagement VLAN now that switch to host has been validated.

Do all the other VLAN settings look correct in PowerConnect with tagging, trunking etc. and Hyper-V configurations with access mode and untagged VM native VLAN?

Also now realizing since the Hyper-V hosts are required to be on the AD domain and the HostManagement will be a different subnet I will need to route at my layer 3 device to be able to contact the domain controllers, that was a missed detail and not ideal since we were looking to keep the HostManagement traffic isolated by way of VLAN like the others.

DELL-Josh Cr · Answer

The other VLAN settings do look correct.

Shawn16 · Answer

I got it working, seemed to be all related to the subnets/VLANS. Can ping back and forth on all (3) VLANS now. I did some test share nothing Live Migrations and it works, but the speed is suspect and before I enable Jumbo frames on the 5548 and the vNICs for LiveMigration wanted to verify everything was correct. The LM throughput avg. 500-700Mbps only. 6Gbps LACP team using (6) NICs from each R710, no weighting set on hyper-V vSwitch either yet. Examining the Team network stats on each server 1 phy NIC on each has a ton more traffic than the other 5 where the 5 are fairly close. One NIC has heavy Rx while the other server has 1 heavy Tx NIC. I'm not even getting a full 1Gb of a single physical NIC and I thought with LACP it was possible to achieve more than the capacity of a single physical NIC. If I'm wrong on that I'm still at a 400Mbps loss of a single phy NIC.

I have not changed any of the powerconnect LACP defaults including priority, timeout is long etc. other than set LAG LB to Layer2/3. When I set LB to layer 2 only I seem to get in the higher range of the 700Mbps throughput so it might perform better than layer 2/3 or just layer 3, but still way below what I would expect. What could be causing my limited bandwidth with teaming? I verified the Task Manager is accurate as it matches my Solarwinds SNMP monitoring.

Shawn16 · Answer

Reason I am not using switch independent (SI) is because we have A LOT of incoming bandwidth usage from streaming video (CCTV) and my understanding is SI will only LB outbound while LACP will do inbound as well, which makes up most of our traffic profile.

I d/l v5.31 of the tool and ran it b/t my 2 new servers and here are some results. Seems with a single thread or multi-thread I am getting similar results just shy of 1Gbps and never anymore. Results below are for an 8 thread transfer, receive then transmit data shown. Results look good to my untrained eye, care to share feedback on the results below?

Also I guess multi-threads still won't get you over 1Gbps throughput, but if I had other TCP sessions going b/t hosts simultaneously I would be able to achieve over 1Gbps?

c:\Temp\NTttcp-v5.31\x64>ntttcp.exe -r -m 8,*,192.168.20.10 -l
Copyright Version 5.31
Network activity progressing...

Thread Time(s) Throughput(KB/s) Avg B / Compl
====== ======= ================ =============
     0   15.013        14630.783     80015.509
     1   15.013        11159.605     80093.411
     2   15.013        14988.591     79292.598
     3   15.013        14510.951     79700.385
     4   15.013        13070.777     79897.069
     5   15.013        14494.995     79301.095
     6   15.013        14690.017     79018.270
     7   15.013        14784.243     79720.373

##### Totals: #####

Bytes(MEG) realtime(s) Avg Frame Size Throughput(MB/s)
================ =========== ============== ================
1646.884491 15.015 1454.894 109.683

Throughput(Buffers/s) Cycles/Byte Buffers
===================== =========== =============
877.461 17.642 13175.076

DPCs(count/s) Pkts(num/DPC) Intr(count/s) Pkts(num/intr)
============= ============= =============== ==============
7152.181 11.053 10475.391 7.546

Packets Sent Packets Received Retransmits Errors Avg. CPU %
============ ================ =========== ====== ==========
371903 1186948 0 0 3.178

-------------------------------------------------------------------------------------------------------------------------------

c:\Temp\NTttcp-v5.31\x64>ntttcp.exe -s -m 8,*,192.168.20.10 -l 128k -a 2 -t 15
Copyright Version 5.31
Network activity progressing...

Thread Time(s) Throughput(KB/s) Avg B / Compl
====== ======= ================ =============
     0   15.016        14627.597    131072.000
     1   15.016        11149.707    131072.000
     2   15.016        14994.140    131072.000
     3   15.016        14508.258    131072.000
     4   15.016        13067.661    131072.000
     5   15.016        14491.209    131072.000
     6   15.016        14678.743    131072.000
     7   15.016        14789.558    131072.000

##### Totals: #####

Bytes(MEG) realtime(s) Avg Frame Size Throughput(MB/s)
================ =========== ============== ================
1646.875000 15.015 1459.962 109.682

Throughput(Buffers/s) Cycles/Byte Buffers
===================== =========== =============
877.456 11.384 13175.000

DPCs(count/s) Pkts(num/DPC) Intr(count/s) Pkts(num/intr)
============= ============= =============== ==============
17647.752 1.403 20959.774 1.182

Packets Sent Packets Received Retransmits Errors Avg. CPU %
============ ================ =========== ====== ==========
1182821 371833 24655 0 2.051

Shawn16 · Answer

I also enabled jumbo frames in my PowerConnect switch and then ONLY enabled jumbo frames on the LiveMigration vNICs to 9014 each since the PowerConnect supports up to 10K. (I didn't think you need to touch the physical NIC attributes since I have a 2012 R2 team) No real benefit in regard to time to migrate a VM or throughput, but I assume if I monitored the switch CPU it likely lessened the load on it (less packets to process).

Networking General

Configuring VLANS Using PowerConnect 5448 & Hyper-V 2012 R2 vNICs

Was this post helpful?