Flings
Apps and tools built by our engineers that are intended to be played with and explored.

VMware Auto Deploy

VMware Auto Deploy

Summary

This Fling is deprecated and no longer updated.

 

VMware Auto Deploy supports automatic PXE boot and customization of large numbers of ESXi systems. Auto Deploy allows rapid deployment and configuration of a large number of ESXi hosts. After a DHCP server has been set up, Auto Deploy PXE boots machines that are turned on with an ESXi image. Auto Deploy then customizes the ESXi systems using host profiles and other information stored on the managing vCenter Server system. You can set up the environment to use different images and different host profiles for different hosts.

System Requirements

VMware Autodeploy requires the following:

  • ESX/ESXi system to install the virtual appliance on. This system must be able to connect to the vCenter Server system on which you want to create and store the host profiles.
  • Static IP address to assign to the VMware Auto Deploy appliance during startup.
  • vCenter Server system to manage the auto‐deployed ESX/ESXi systems and to store the host profiles.
  • MAC address or asset tag of each ESXi system you want to PXE boot.
  • DNS Server that maps IP addresses to host names. (Optional)

Instructions

Quickstart Instructions

  1. Deploy the OVA file on an ESX/ESXi server. The autodeploy VM should have a NIC connected to the management network.
  2. Boot the new VM.
  3. Go to the VM's console and answer any questions.
  4. If the management network has no DHCP server, login as vi-admin and run 'sudo deploy-cmd dhcpconfig'. Otherwise, consult the manual for setting up the DHCP server.
  5. Network boot the machine to be used as an ESXi server. It should contact autodeploy and start booting ESXi.

Please refer to VMware Auto Deploy Administrator's Guide for more details.

Video

Change Log

Engineers


Timothy Stack

Management Platform

Greg Hutchins

Management Platform

Patrick Devine

Management Platform
Add Feedback

46 thoughts on “VMware Auto Deploy

  1. talazac

    I have seen on the appliance, all is installed by RPM.
    Typical in the rpm vmware-deploy…..
    It is possible to have this RPM to configure my red-hat installation. And not duplicate an appliance.

    Thanks in advance

  2. Robert

    I’d like to setup Autodeploy in a large nertwork in which there is already a DHCP server and a PXE server. Would I be able to setup Autodeploy using an existing PXE server?

    Thank you

  3. kirk

    Are there any examples of how to change the name of the new ESX host without relying on the DNS resolution? In my case DNS may not always be available.

  4. kirk

    I have tested the auto deploy against vcenter 5.0 beta and found it to to work, with some issues.

    1. A configured VMAD os profile with a folder definition does not put the new ESX host in that folder as documented
    2. The new ESX host (added as a standalone host by VMAD) has communication issues in VCenter 5.0.
    3. A failed DNS resolution for the new standalone esx server causes the vcenter server to crash.
    4. The host profile does not get applied to the new host due to communication issues.The deploy log shows the host profile is being applied and that it fails for a communication issue.

  5. Kipz

    hello,

    has anyone tried AutoDeploy with Cisco UCS B-series blades? i’m trying to get it working and no success so far. my blades are using M81KR virtual interface CNA-s. blade is able to contact DHCP server and loads successfully gPXE but then gets stuck with second request to DHCP for interface “net0”.

    the same is from gPXE command line. when i manually issue command “dhcp net0” then it stucks there:

    gPXE> dhcp net0
    DHCP (net0 00:25:b5:00:00:2e).

    i never get that “… ok” what should indicate successful DHCP request. and it stays there forever, does not get even timeout. from DHCP server tcpdump i see that actual DHCP request reached server and server sent out the response, but seems that client does not get that response or is not able to use it. my DHCP server is Windows 2003 configured according AutoDeploy manual.

    i’m able to successfully PXE-boot other servers like Cisco C-series rack-mount machines or even virtual machines. this indicates that DHCP server works correctly. but none of my blades succeeds…

    any ideas?

    regards,

    kipz

          1. John Kennedy

            So I found out that the AutoDeploy in vSphere 5 works with the m81kr I have installed, but not the m71KR-q (older Menlo card). M71KR-1 has an Intel 82598EB 10 gig Ethernet card on board, which does not work with the version of pxe that AutoDeploy uses. That’s all I know so far. If I get hold of a M72KR, I’ll test it.

            1. T.J. Albrecht

              What settings are you using for this to work? When I run UCS with M81KR PALO cards, the gPXE fails on the second DHCP attempt for booting.

              I have tried vyatta DHCP server, Windows DHCP server (2008 R2), and TFTPd32. All with no luck. Are there specific options that need to be passed or utilized? Or does gPXE 1.0.0 need driver updates for the PALO to work?

              1. Àngel Casanova

                The same problem here.
                I’m using the M81KR with vSphere 5 Auto Deploy and the second dhcp request fails.
                In fact, the second DHCP discover arrives to the DHCP server (so the nic is working inside the gPXE environment) and the DHCP Server answers with a proper DHCP response that never seems to arrive to gPXE…

                1. T.J. Albrecht

                  I never saw a second OFFER my from my DHCP server. I see the second DISCOVER from the gPXE client, but no response OFFER. What DHCP server are you using and what settings? I tried vyatta, windows 2008 r2, and tftpd32.

                  1. J. Lukes

                    I spent several days troubleshooting the UCS/M81KR issue with Cisco TAC. They’ve managed to reproduce the problem in their labs after I provided packet captures and some exhaustive troubleshooting. We’ve determined that it is without a doubt a incompatibility between the M81KR cards and Windows DHCP. Currently I am using Linux DHCP as a workaround until they fix this bug. CSCts86689

    1. Tim Stack

      gPXE needs a change to deal with the NIC. Basically, the NIC reports that it uses interrupt zero to notify gPXE that a new packet has arrived. However, interrupt zero is the timer interrupt and that occurs too infrequently for it to be useful. So, in reality, gPXE needs to poll the NIC for packets instead of waiting for the interrupt.

      You can probably use the latest version of iPXE with the change to report a different syslinux version number.

  6. Jonathan Sabo

    You should use Cobbler instead of this… It’s so annoying to have to manage your virtual infrastructure with different tools than your physical infrastructure, especially when you dont have to. Cobbler can install ESX and ESXi…. and create vms too!

    1. maddy04

      Indeed. But we have host profile working nice and cool in auto deploy and it’s good for Enterprise environment where they want to make bulk changes often. we like it !

  7. GavinJB

    This has been working marvellously within my lab infrastructure until I recently updated the default boot profile to use a new image (ESXi 4.1_Update1-348481) – which now reulsts in a PSOD at the Reconnecting to VC Stage…

  8. Reid Loper

    We’re having issues getting the Linux vCenter virtual appliance Auto Deploy to function in our test environment. The lab consists of HP BL460c blade systems with NC553i integrated network cards.

    I started with an out of the box configuration using the undionly.kpxe boot file, and DHCP and TFTP servers configured on the appliance.

    1) The BL460c booted and received an IP address. It was directed properly to the TFTP server where it downloaded the undionly.kpxe file.

    2) Using the undionly.kpxe file the host made a second request for an IP address from the DHCP server, but timed out without receiving an address.

    Investigating the issue, I learned that gpxe is no longer actively updated by its community. Suspecting that the issue could be driver related we swapped the gpxe-based undionly.kpxe file with a newer ipxe-based undionly.kpxe file and the booting host was able to receive an IP address when it made the second request.

    However, a new issue appeared. The host was unable to make a HTTP connection to the auto deploy server to download and execute the http:///vmw/rbd/tramp file. SSL is used by the auto deploy server and I attempted to change the URL to “https”. However, iPXE does not support HTTPS. I eventually turned off SSL on the remote boot virtual server on the appliance and while the booting host could then connect to the auto deploy server using http and the port 6502 of the virtual server, it could not execute the tramp file.

    Does the VMware distribution of gPXE undionly.kpxe have any VMware customizations that make it have to be used for the pxe boot? Is there any update to the file or is there a way to inject updated drivers into the gPXE boot file?

  9. webbed7

    Hey,

    We are not licenced to use “Host Profile”, can i still use auto deploy to provision multiple esxi 4.1 hosts ? we are licenced for just Ent and not Ent Plus. I read the admin doc, it talks about host profile etc.

    Thank You

    1. Tim

      Host profiles takes care of the configuration so, without it, you’ll need to find a way to configure the hosts every time they boot up.

  10. Paul Wood

    I have this working fine using the DHCP server to add the host and ip address. My issue is the hostfolder option. It does not seem to matter what I setup it never puts it in the cluster I want it to use.

    The command is deploy-cmd updateprofile -n -f /DC/Cluster/. The issue is it just adds to the DC and not to the cluster.

    Any ideas ?

  11. Chris Paggen

    My hosts are able to boot using the autodeploy appliance. They just never get inserted into vcenter unless I reboot the autodeploy appliance (!). Hosts seems to remain stuck in BOOTING status and the logs show nothing extraordinary:

    Jan 5 04:59:09 autodeploy deploy-cgi[2291]: automatic addhost — {‘assettag’: ‘ Unknown’, ‘profile’: ‘default’, ‘status’: ”, ‘bootmac’: ’00:25:B5:58:10:9F’, ‘l ast_booted’: None}
    Jan 5 04:59:09 autodeploy deploy-cgi[2291]: host boot — {‘profile’: ‘default’, ‘status’: ”, ‘last_booted’: None, ‘assettag’: ‘Unknown’, ‘id’: 2, ‘bootmac’: ‘ 00:25:B5:58:10:9F’}
    Jan 5 04:59:41 autodeploy deploy-cgi[2291]: reverse lookup failed for 192.168.9 9.198, using IP address as name in vCenter
    Jan 5 05:03:11 autodeploy deploy-cli: reverse lookup failed for 192.168.99.198, using IP address as name in vCenter
    Jan 5 05:03:11 autodeploy deploy-cli: starting escort of host 2 (192.168.99.198 ) — host(2, ’00:25:B5:58:10:9F’, ‘192.168.99.198’, ‘Unknown’, ‘default’, ‘2011- 01-05 09:59:09’, ‘BOOTING’, ‘2011-01-05 09:59:09’), boot-profile(‘default’, ‘192 .168.99.99’, ”, ‘b.z,k.z,s.z,c.z,oem.tgz,m.z’, ‘no-auto-partition’, ‘stargate_p rofile’, ”, ”)

    [vi-admin@autodeploy ~]$ deploy-cmd listhosts
    Host ID: 2
    Boot MAC : 00:25:B5:58:10:9F
    IP Address : 192.168.99.198
    Asset Tag : Unknown
    Boot Profile: default
    Status : BOOTING (last booted: 2011-01-05 09:59:09)
    [vi-admin@autodeploy ~]$

    I’m guessing some race condition is preventing the SDK script from firing up.

    1. Tim Stack

      The appliance is just trying to contact the host after the “starting escort of host” message. It seems strange that it would not be able to connect. Is it logging any messages like “waiting N more seconds for hostd to spin up” ?

      You can manually run the command to add the host to vcenter by running “deploy-cmd escorthost -i “. You might want to play with that when debugging.

  12. Cedric Laurent

    Hello,

    We have 2 ESX 4.1 i Hosts VMs wich boot from the VMware Auto deploy PXE Server,
    how can i bind 2 host’s Ethernet MAC address to an IP address on the /etc/dhcp.conf.template file.

    Thank you very much

    Bye Bye

    Cedric

    1. Tim Stack

      Something like:

      # Assign a static IP to foo
      host foo {
      hardware ethernet 00:45:40:10:FE:12;
      fixed-address 10.1.1.20;
      }

  13. Michael Wilmsen

    Hi,

    Nice job! But is there a way to get ESXi installed on the local disk, because now if I reboot the host without PXE I get a no operating system found. It seems that the boot of ESXi is just in memory. I tried to turn of the no-auto-partition (entering –bootoptions=””) but this does not work.

    Thanks.

    1. Tim Stack

      You’ll need to add the installer to the list of modules to boot and pass a “ks=” boot option to specify the kickstart file. I think there are instructions available for doing PXE installs of ESXi. You just need to do some minor adaptation when using autodeploy.

  14. juerg rietmann

    I’m about to install the Auto Deploy appliance on my ESX Server. I need to assign a static IP to the appliance, thus edit the ifcfg-eth0 file. How do I get the appropriate rights ?
    Cheers .. juerg

  15. Robert Stothard

    I am attempting to set up dhcp with command string deploy-cmd dhcpconfig I have logged in to ssh with viadmin however I recieve the message You must be root to execute this command. Also at the console the Alt-F2 key stroke does not open interface.

    Thankyou

  16. Mike

    This looks like an interesting product. Is Enterprise Plus licensing required. We have Enterprise and I know host profiles only come with Enterprise Plus.

  17. Duco Jaspars

    Gerat tool!
    What is the best way to provide feedback on Auto Deploy? Feel free to drop me a mail if needed.
    Regards, Duco

  18. NiTRo

    Hi guys !
    Great works on that one, is it supported in production yet ?
    Where can i found others plugin samples or something to help on this ?

    1. Tim

      It is not supported in production.

      I don’t think we have any other plugin samples right now other than what is in the docs. The AutoDeploy python code is in the virtual appliance, so you can mess around with it directly.

    2. sdo@bm

      @Nitro
      the buttom “I agree to the TECNICAL PREVIEW bla bla bla” should be a hint that this is NOT officially supported by your license agreement and as such cant expect VMware to help you if it trashes you infrastructure .. play around with it in test environments and learn the feel of it till the final release ..

Comments are closed.