3 Replies Latest reply on Feb 4, 2019 6:10 AM by larstr

    network performance problem in CentOS VMs after Migrationg Network from port Group to Logical Switch in NSX

    emmoon1981 Novice

      Hello

       

       

      I Have problem in network performance ( network speed) after migration network infrastructure from port group to logical switch in VMs has Linux OS (Cent OS Distribution ) in NSX for example Copynig file From/To those VMs have Cent OS  is too low .This problem ( Low Speed coping file in network to VMs) just for Cent OS VMs and other VMs has Microsoft OS doesnt have any problem of copying file from network.

       

      Thanks for your help.

       

      Regards,

      Ehsan

        • 1. Re: network performance problem in CentOS VMs after Migrationg Network from port Group to Logical Switch in NSX
          sk84 Expert
          vExpert

          Which NSX and vSphere version? Which network adapter is used by your VMs (vmxnet3, E1000 or whatever)? How are the VMs connected? And from which source do you copy? Or, more precisely, are any firewalls, edge gateways or DLR in between? Has anything changed at the storage level? What physical bandwidth capacity is available?

           

          Can you show with examples how slow the copy process over the network is? How much throughput do you get with CentOS VMs and how much with Windows VMs?

           

          What happens if you copy a file from CentOS VM 1 to CentOS VM 2 and both VMs are on the same logical switch?

           

          Can you rule out that there is a limit on the physical network or that the problem only occurs on one ESXi host? Are there configuration differences between the logical switches for Windows or CentOS? Or do you use the same logical switch for all VMs in your test?

           

          To help you, we need to understand exactly what your setup looks like. So the more information you give us, the more likely we are able to help you. ;-)

          • 2. Re: network performance problem in CentOS VMs after Migrationg Network from port Group to Logical Switch in NSX
            emmoon1981 Novice

            Hello

            1. Which NSX version is used? 6.3.5.7119875
            2. Which vSphere version is used? vCenter:6.5.0 ,    ESXi:6.0.0   Build3620759
            3. Which network adapter is used by your VMs (vmxnet3, E1000 or whatever)? We have same problem with VMXNET3 and E1000.
            4. How are the VMs connected? Two VMs connected through Nexus 5000 and there are any firewalls or edge services between them. We have a DLR for routing between logical switches, but same problem is remained with and without DLR.
            5. Has anything changed at the storage level? No
            6. What physical bandwidth capacity is available? 2*10G CNA with Active/Passive scenario connected to Nexus 5000
            7. How much throughput do you get with CentOS VMs? I got 12 Mbit/s bandwidth with scp command in two  Linux VMs. Two VMs are in a logical switch.
            8. How much throughput do you get with Windows VMs? I got about 400 Mbit/s with two Windows VMs.
            9. Can you rule out that there is a limit on the physical network or that the problem only occurs on one ESXi host? There is any limitation on physical network and we have the same problem on one and two hosts.
            10. Are there configuration differences between the logical switches for Windows or CentOS? I use the same logical switch for both of them.
            11. I have a serious Problem: I have a DLR that connected to a logical switch (vxlan protocol) and a vlan (on distributed switch). When I transfer a file between two Linux VMs in a logical switch, I get 12 Mbit/s bandwidth limitation .I get the same problem when I copy a file with scp command when a VM is in logical switch and another VM is in vlan network (on distributed switch). But when a Linux VM is in vlan and another VM is in logical switch that both of them connected to a DLR, I can’t copy a file with scp command from vlan to logical switch (vxlan), but there is ping between them!!!!

            Regards

            Ehsan

            • 3. Re: network performance problem in CentOS VMs after Migrationg Network from port Group to Logical Switch in NSX
              larstr Virtuoso
              vExpert

              Eshan,

              You have connected it to some logical network and a vDS portgroup? Normally you should have only logical switches on your DLR and communication with other networks should go through an Edge gateway.

              Other than that you should also check the following:

              1. Is your Centos VM having multiple interfaces? Is routing enabled?

              Poor network performance when using VMXNET3 adapter for routing in a Linux guest operating system (2077393)

              2. Is performance bad when using both TCP and UDP?

              Poor TCP performance might occur in Linux virtual machines with LRO enabled (1027511)

              3. Are your problems related to IPv6?

              System using vmxnet3 NICs discards some IPv6 traffic and performance is poor

               

              Lars