ESXi install success, but no access via management IP

[nscott]

I've installed ESXi v4 onto an AMD 64 machine with a Intel PRO 1000GT card. ESXi 3.5 worked fine with these cards and it appears that many people have gotten this card to work on their installs of v4.

While ESXi v4 installed fine, it won't pull a DHCP address which is not an issue for any other machine on the network. Even more strange is that if I set a static address it becomes reachable from other machines on the network (ping test). The problem is, that while ESXi appears to be working and ready for a connection with the "Download tools to manage this host from: (STATIC)" message - it is not accessible from a browser anywhere on my network. I can ping it, but I can't reach the address via http in a web browser.

The network test which pings the DNS servers and gateway all come back as successful so the machine appears to be able to ping outward as well.

Any ideas?

Thanks,

Nate

[Dave_Mishchenko]

Welcome to the VMware Community forums. Do you have any other NICs in the server? If you access the console and run esxcfg-nics -l does the speed and duplex match what you expect the switch to be set to?

[Randall Robertson]

Can you use the local interface to look at the logs? You should see messages from dhclient about why it rejected an address (if it got one) it could be that your DHCP server sends out invalid responses and our client is stricter than others on your network.

As to why you can ping the static address, but not reach over http, I am not sure. Perhaps that static IP isn't really unused, and someone else on the network has it, so you are hitting them. This, (combined with you having 2 nic ports, only one of which is plugged in, and ESX chooses the one that isn't plugged in) would explain both a lack of DHCP working an an apparent ability to ping but not to get HTTP port 80 traffic.

Do you have a LInux host (or can you use a LiveCD) to run nmap against the ESXi IP address and validate which ports are open?

You could also try to 'Restart Management Agents' once you have set the static IP, although this should not be necessary.

[nscott]

I'm working within a very small network so it's pretty easy to assertain whether the IP address is available or not. If I start pinging the address prior to setting it in the static IP settings it is unreachable. As soon as I hit "Yes" to the "apply changes and restart management interface" prompt in ESXi it becomes available and begins responding (and I verified by NIC lights that it is indeed this host responding). I did some further digging and in the logs I'm getting some errors related to the HTTP request. When I try to connect to it straight from a VI client the logs show an error in the SSL handshake. Unfortunately, I ran out of time to get this working and had to go back to another Virtual host solution on this machine so I can't give you the exact errors. It's a separate hard drive so I hope to have some time to give it another shot in the next couple of days.

I did go ahead and install a full ESX (vSphere) 4 install as well and it works fine that way in that I can get to the server via HTTP, it picks up a DHCP address, and the VI client can access the host just fine. This is on the exact same box with the exact same Intel PRO 1000GT NIC which makes the issues with ESXi seem even more strange. The problem I have on ESX vSphere is that I can't access the VM console of any VM that I setup - cannot access MKS, timeout error. I'd guess this is a networking error (I did try another NIC just to test and same issue).

In both cases, I have no DNS server active on the network. My DNS is running within a VM and I currently only have the one host to work with (very small business here) so it's a bit of a challenge to determine the role of the DNS service in all of this. I have hard coded the IP address in the hosts file of the host and any client trying to reach the server but it wouldn't be the first time DNS was an issue with VMware for me. I'll build a physical DNS server and try from there this week as well. I have an ESXi 3.5 u3 server at my other office so I'll put them on the same physical network and try again.

I thought at one point this may be an issue with a client side firewall so I did boot an XP machine with no firewall installed just to try and connect to the host. Same issue. I do have a Linux box running on my 3.5 host as well so I can try finding the open ports on the ESXi host as one of you suggested as well. The interesting thing is that my web client never comes back with an error when trying to reach ESXi on the HTTP address. It's like it is just waiting for a response - never gives a host not found or unreachable error. VI client gives a time out error, not a host not found error as well.

Thanks for your help.

[sooth]

Here I finally found someone getting exact same error as I am. Brand new (unsupported) Gigabyte M61PME-S2P, dual core 64-bit Athlon 7750, 4Gb, (supported) Intel PWLA8391GT 10/ 100/ 1000Mbps PCI PRO/1000 GT. DHCP fails. So I configure static IP. Ping from client box to ESXi succeeds. Ping from ESXi to client box succeeds. Cannot get a response from ESXi from the http request. Probably sunk 12 hours. Losing hope.

[sooth]

Hello, sorry, more details: ESXi vers. 4.0.0 build 171294. To confirm the hardware is ok I installed ubuntu server and everything operates fine (DCHP worked, static IP worked, and apache2 fielded http requests just fine). Have tried installing ESXi 4.0 to the 250Gb SATA drive and have also tried installing to a 2Gb USB drive. Get stuck at the same point every time. The given http:// address in the banner screen does not respond to IE8 on vista ultimate or IE7 on XP Home or Firefox on ubuntu 9.04 workstation. FYI and thank you.

[sooth]

If I Alt-F1 into the ESXi unsupported shell, I can wget the given address and indeed wind up pulling the index.html file. From anywhere off the ESXi host, the http request goes nowhere.

I pressed Alt-F12 and looked at the "vmkernel log". There is a line in reverse video that says

0:00:00.834 cpu0:4798)WARNING: Tcpip_Socket: 1046: failed to unset the ip address (error=49)

[edurrant1]

Hi, I have just created a similay (AMD Phenom) system and after getting the "lvmdriver" error bought a cheap Intel Pro/1000 NIC - in my case the MT not the GT and that one works fine with DHCP and comunicates without issues to the other systems on my home LAN. So I would suggest this could be a driver problem on the Intel Pro/1000GT driver?

Another point that may help someone - my finding is that the VSphere management client will NOT successfully install under Windows 7 RC. I have had to install it in the Virtualised Windows XP environment under Windows 7. It's a dog to install - it takes forever!

Now to see if I can get the ESXi to recognise my IDE DVD-RW drive so that I can install a virtual machine ....

Cheers/2

Ed.

[sooth]

Hello edurrant1, thank you very much for the advice, driver issues would certainly explain my trouble.

I found this post which seems like exactly what I need.

In light of the fact that the currently installed drivers are impairing the network, how can I update the drivers on the ESXi host? I have downloaded the updated drivers .ISO, burned onto CD, but now what?

Thank you again

[edurrant1]

Hi,

I am new to using ESXi, but reading a link from the URL you gave, I think the following text is relevant, as you wish to add/update a driver to an existing installation:

(for both ESX and ESXi)

1. Power on the ESX or ESXi host.

2. Place the driver CD in the CD-ROM drive of the host where either the vSphere CLI package is installed or vMA is hosted.

3. Mount the driver CD.

4. Navigate to <cd mount point>/offline-bundle/ and locate the INT-intel-lad-ddk-igb-1.3.19.12-offline_bundle-166506.zip file.

5. Run the vihostupdate command to install drivers using the offline bundle.
   vihostupdate <conn_options> --install --bundle INT-intel-lad-ddk-igb-1.3.19.12-offline_bundle-166506.zip

For more details on vihostupdate, see the [vSphere Command-Line Interface

Installation and Reference Guide|http://www.vmware.com/pdf/vsphere4/r40/vsp_40_vcli.pdf].

[sooth]

Hello again edurrant1, thank you very much for your help. Problem is, I haven't yet gotten the client downloaded (since the network doesn't work) and even if I did, I am not sure I could update ESXi from the client (ditto).

I squinted at the HCL and now realize that the card that I paid $30 and waited a week for in order to supplant the perfectly operational on-board ethernet port is actually Not Supported.

I have been reading about Ubuntu KVM. It doesn't make any sexy "bare metal" claims, but it does all processor state stuff in hardware, and uses the very stable and optimized linux kernel for memory mgmt and scheduling. AND you get all the hardware support and up to date drivers via your linux platform. It's probably all command line without any charts and graphs but god this hardware / driver / HCL stuff is ridiculous.

[edurrant1]

I think (and I could have misunderstood this) that the commands to install the driver can be executed in the console on the ESXi machine - if not as you say, the process wont work. In that case, looking again at the link you sent there are instructions how to include the new driver during the initial install of ESXi - so one other option would be to re-install ESXi and include the revised driver at install time. as you wont have created any virtual machines yet, you wont lose any work.

The VMWare hypervisor (ESX) has always run only on a restricted amount of hardware types. ESXi v4 supports a LOT more hardware than earlier versions. It's really meant to run on server equipment certified to run it. I am actually surprised it works on my hardware.

I'm sure you know this but just in case others don't: There are other VMWare virtualisation engines that run on top of Host OSes (also from Microsoft, Sun, Citrix and Open Source) but these are dependent upon the Host OS (Windows, Linux, Mac OSX, OS/2-eComStation etc.) - if the host OS fails or needs patching, all your virtualised systems get closed down. The host OS in these cases also takes a big "chunk" of available resources. However running on top of a host OS means that you don't have to worry about hardware compatibility with the virtualisation engine as it talks to the host OS not the physical hardware. Some of these packages now take advantage of the chip level virtualisation features to speed their operation.

I looked for a network card after I got the "lvmdriver cannot load error" and indeed having looked at the forums went searching for an Intel Pro/1000 GT card but when I found them costing between AU$90 and AU$250, I decided I'd risk buying the AU$22 Intel Pro/1000 MT card instead, as if it didn't work, I could always use it for another system. I was lucky - it worked straight away.

Cheers/2

Ed.