Advertise on Bikeforums.net



User Tag List

Results 1 to 6 of 6
  1. #1
    derailleurs are overrated bigbenaugust's Avatar
    Join Date
    Feb 2005
    Location
    KIGX
    My Bikes
    2009 Motobecane Fantom CX, 2012 Motobecane Fantom Cross Uno SSCX
    Posts
    1,720
    Mentioned
    1 Post(s)
    Tagged
    0 Thread(s)

    Any VMware guys in here?

    It's Thanksgiving, and yet work is preying upon my mind.

    I took our two VMWare Server 2 (on RHEL5) machines and reinstalled them with ESXi 4.1u2. I was also going to do a P2V conversion of one more WS2k8 server, but that has been prevented due to a little networking issue that I am going to describe below.

    The ESXi host is configured thus:
    (A and B are WS2k3, C is WS2k8)

    Code:
    VM A---|vswitch1--pnic1--network 1 (X.Y.Z.Q)-(real wire)-(database network)
    VM B--/
    
    VM A---|vswitch0--pnic0--network 2 (192.168.1.Z)-(real wire)-(non-routed internal subnet)
    VM B--/
    
    VM C---|vswitch2--pnic2--network 3 -(real wire)-(desktop network)
    The ESXi default gateway is on network 1. VM A and VM B are pretty happy. They aren't publishing to my BackupExec server over network 2 like they used to, but they can ping hosts on both networks 1 and 2 and are doing their server functions over network 2 (as they should be) just fine for now. They're FileMaker machines living out their autumn years virtually before I terminate them with extreme prejudice in the coming months.

    I have no trouble accessing the machine via vSphere Client over network 2, but at some point it quit working over network 1.

    But VM C (my P2V conversion victim) can't get out and no one can see it. It essentially has no network. The VMware networking docs I read said I wouldn't need to give the machine an address on network 3 unless I wanted management access. (It doesn't matter, it doesn't work either way.) I know the interface is getting traffic, I can see the blinking lights.

    I can't set static routes in ESXi without interfering with the automatic routes set by the machine. ESXi only takes one default gateway per machine, so I can't really do it on a per-interface basis.

    Do I need a virtual routing appliance to fix my traffic problems inside the machine? Or is my VMware setup fine and there some stupidity inherent to WS2k8 that I am missing here?

    Thanks for listening, y'all.
    --Ben
    Carrboro Bike Coalition - putting the "bike" in "CARrboro" :)
    2011 Motobecane Fantom Cross Uno, 2009 Motobecane Fantom CX
    Previously: 2000 Trek 4500 (2000-2003), 2003 Novara Randonee (2003-2006), 2003 Giant Rainier (2003-2008), 2005 Xootr Swift (2005-2007), 2007 Nashbar 1x9 (2007-2011), 2011 Windsor Shetland (2011-2014)
    Current Linux Usage (by machine): Arch: I Debian: I openSUSE: II

  2. #2
    Senior Member DannoXYZ's Avatar
    Join Date
    Jul 2005
    Location
    Saratoga, CA
    Posts
    11,495
    Mentioned
    0 Post(s)
    Tagged
    0 Thread(s)
    Couple questions:

    1. how did you migrate the VMs from Server2 to ESXi?
    2. is VMwareTools for ESXi installed on all three VMs and running properly?
    3. what is subnet IP, mask and gateway on Network-1? Subnet IP & mask on Network-2? Network-3?
    4. what is IP/mask/gateway of each NIC on VM-A, B & C?
    5. are all three pNICs plugged into same external physical switch?
    6. how many external physical switches? VLANs?
    7. what type of NIC Adapter Type is configured for VM-C? E1000? VMXNET3?


    Typically issues like this are due to mask & gateway settings of the OS inside the VM. To distinguish between VM settings versus OS settings, set VM-C's NIC to network 1 or 2 and give it IP in same subnet and see if it can ping VM-A or B on those networks. If it can, then OS settings are fine, it's a VM configuration problem.

    I've also seen this issue with VM-C where the virtual-NIC of the VM is disconnected. Verify that both Connected and ConnectedAtPowerOn are checked in the VM settings.

  3. #3
    derailleurs are overrated bigbenaugust's Avatar
    Join Date
    Feb 2005
    Location
    KIGX
    My Bikes
    2009 Motobecane Fantom CX, 2012 Motobecane Fantom Cross Uno SSCX
    Posts
    1,720
    Mentioned
    1 Post(s)
    Tagged
    0 Thread(s)
    This is why I love Foo.

    1. vCenter Converter Standalone in all cases. VM A & VM B were copied off of the VMWare Server 2 machine and run through Converter to put them on the ESXi machine. VM C was converted from a physical machine to a VMWare Server 2 VM, then to the ESXi server.
    2. Yes and yes.
    3. Network 1: 171.67.X.144/255.255.255.240, so gateway is 171.67.X.145. (default gateway in ESXi is on this net)
    Network 2: 192.168.1.0/24, there is no gateway. Pretty much 99% of the traffic flows from machines to the backup server on 192.168.1.22. There is also some SSH/SCP/Rsync traffic between machines that goes on there, too.
    Network 3: 171.64.Y.0/23 (the desktop net)
    4. For A & B, all of that info is on my desk or on VPN. I can come up with it later. VM C is set up like a normal desktop for us, addr: 171.64.(Y+1).147, nm: 255.255.254.0, gw: 171.64.Y.1.
    5. We have two switches in the server room (coupled together, 96 ports) that are VLANed out into 7 different networks (web, db, servers, other, desktop, the internal, and one net for a different group). So each plug is on a different VLAN on the same physical switch. I figured if I had done anything stupid, Networking would have given me a call (like the time I plugged in a Cisco load balancer and they thought I was plugging in an unauthorized router).
    6. see above.
    7. As converted, VM C had a VMNet AMD adapter. I removed it and changed the VM to have a E1000 instead. It didn't work either way. Physical card is a Broadcom (uses tg3 driver) (the other two physical interfaces are e1000).

    I did check the Auto Connect box, but it does get set in a couple of different places, so I could have missed it the 15th time (should have gone to Foo earlier!). I was considering re-IPing VM C (the least destructive option) to network 1 on Monday. I was also considering getting out my Debian installer CD and creating a tiny VM on network 3 just to see.

    I suppose I could get Networking to run all three networks on the one port and use VLAN tagging or something, but I do not think that is necessary.
    --Ben
    Carrboro Bike Coalition - putting the "bike" in "CARrboro" :)
    2011 Motobecane Fantom Cross Uno, 2009 Motobecane Fantom CX
    Previously: 2000 Trek 4500 (2000-2003), 2003 Novara Randonee (2003-2006), 2003 Giant Rainier (2003-2008), 2005 Xootr Swift (2005-2007), 2007 Nashbar 1x9 (2007-2011), 2011 Windsor Shetland (2011-2014)
    Current Linux Usage (by machine): Arch: I Debian: I openSUSE: II

  4. #4
    Senior Member DannoXYZ's Avatar
    Join Date
    Jul 2005
    Location
    Saratoga, CA
    Posts
    11,495
    Mentioned
    0 Post(s)
    Tagged
    0 Thread(s)
    Ok, let's verify that network-3 on the host is connected to a properly VLANed physical switch port.
    1. enable SSH on ESXi host: Using Tech Support Mode in ESXi 4.1
    2. use Putty to SSH to the host, enter root credentials
    3. ping gateway on network-1, ping server on network-1, both should return ping times
    4. ping server on network-2, like backup-server 192.168.1.22, should be successful
    5. ping a known working desktop on network-3
    6. ping gateway on network-3, 171.64.Y.1, is this a real physical port? On a router? Switch?

    If step 5 & 6 are success and return ping times, the networking on the ESXi host is correct and we have a VM configuration issue. We should clean up ALL old drivers and hardware settings on VM-C that was carried over from the physical box.
    1. Log onto VM-C with admin account
    2. bring up command-prompt, type in "set devmgr_show_nonpresent_devices=1 <enter>"
    3. then "devmgmt.msc <enter>"
    4. when the DeviceManager comes up, pull down from menu-bar: View -> Show hidden devices
    5. expand each section by clicking the + and inspect the devices listed
    6. you'll find a tonne of greyed-out entries that represent the old hardware from Server2 and the physical box you virtualized
    7. highlight each greyed-out phantom device and hit DEL key, then click OK to uninstall and remove the old drivers

    Most problematic in this case is the old network-drivers from the old box. Windows will still load the drivers on boot and the original IP will still be bound to the old NIC. This can cause problems when you try to assign that same IP to the new E1000 virtual-NIC. Once all the phantom devices have been removed, reboot. During the reboot, verify in VM settings that the E1000 NIC is assigned to network-3 and that both Connect and ConnectAtPowerOn options are checked.

    After VM-C reboots, login with admin-account and go to control-panels-> network connections. View -> Details and you should see the following:
    status = enabled
    device name = Intel(R) PRO/1000 MT Network Connection
    type = LAN or High-Speed Internet
    name = name of your workgroup or Domain if you have DC on network-3

    Now it's just a matter of assigning proper IP and mask for network-3 and VM-C should be on that network.
    Last edited by DannoXYZ; 11-25-11 at 06:46 PM.

  5. #5
    derailleurs are overrated bigbenaugust's Avatar
    Join Date
    Feb 2005
    Location
    KIGX
    My Bikes
    2009 Motobecane Fantom CX, 2012 Motobecane Fantom Cross Uno SSCX
    Posts
    1,720
    Mentioned
    1 Post(s)
    Tagged
    0 Thread(s)
    I didn't know you could make the old devices show up to delete them. On the WS2k3 machines, I did get complaints about the new interfaces having the same addresses as the old ones. WS2k8 didn't object in the same way. I will definitely look into that.

    The network setup didn't seem to be terribly esoteric based on my research, so I was wondering if it was a Windows issue.
    --Ben
    Carrboro Bike Coalition - putting the "bike" in "CARrboro" :)
    2011 Motobecane Fantom Cross Uno, 2009 Motobecane Fantom CX
    Previously: 2000 Trek 4500 (2000-2003), 2003 Novara Randonee (2003-2006), 2003 Giant Rainier (2003-2008), 2005 Xootr Swift (2005-2007), 2007 Nashbar 1x9 (2007-2011), 2011 Windsor Shetland (2011-2014)
    Current Linux Usage (by machine): Arch: I Debian: I openSUSE: II

  6. #6
    Senior Member DannoXYZ's Avatar
    Join Date
    Jul 2005
    Location
    Saratoga, CA
    Posts
    11,495
    Mentioned
    0 Post(s)
    Tagged
    0 Thread(s)
    Oh, there needs to be VLAN configuration as well. It can be done in one of three ways:
    1. External Switch Tagging (EST)
    2. Virtual Switch Tagging (VST)
    3. Virtual Guest Tagging (VGT)

    Using #2 or #3 requires the physical switch-port to be configured as trunk. Most likely it's an access-port if it is used for physical-servers. So you can only use method #1 External Switch Tagging with VM-C.

    Here's a good article on ESX/ESXi connecting to physical switch via VLAN access mode and External Switch VLAN Tagging (EST Mode)
    Also: ESX3-VLAN paper.pdf

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •