ESX Deployment Appliance (EDA) is a small and easy to use appliance
that makes deploying ESX servers a breeze. It has a very intuitive
web-interface that can configure and deploy dozens of ESX servers in
minutes. It has a script-builder that will allow any admin to create
%post-scripts that will do most anything one needs to get the ESX hosts
up and running! Even if deploying with RDP/Altiris or the UDA, this
script-builder can help setting those up very quickly.
sounds like the install finishes correctly but the resulting server has no connectivity? that should not be possible :smileysilly:
and in the latest version there's an option to add lines yourselve. in the next version i hope to have an import/export function and a way to order them. that way i can setup a place on the internet where people can get (and upload hopefully) specific scriptparts. updating and extending the scriptbuilder should be real easy then.
for now, the scriptparts are also backed-up so you anything you add will be in the backupfile. that way when a new release comes out, you just restore it and your personal additions will be available to you again.
Thanks for the response. However, I don't believe this is a networking issue since the only change is EDA version. With EDA 0.81 using ESX 3.5U3 and keeping the same configuration for the same hardware, I was able to install ESX 3.5U3 without any issue. When I change to EDA 0.85 and everything else stays the same, I get the message about selecting the NIC to continue. So if the EDA is the only variable that changed and I am using the same server and same ESX ISO, the network is not the problem. In addition, i can constantly reproduce this problem with other blades in the chassis with EDA 0.85
I ran into a similar problem and tracked it down to the network trunk configuration. In my case I was using Network Trunks with a native VLAN of 21. In my service Console Port group I had set it to use VLAN21 as well, however native VLANs should not be tagged. As soon as I cleared the VLAN on the Service Console Port Group it worked fine.
STP Can affect this because on Cisco devices it places the port in listening mode when it first comes up. It can take up to 30 seconds for the port to go from Listening to Learning and on to Forwarding mode. Once it is in Forwarding mode it will begin passing traffic, prior to this no traffic is passed on the port. The kickstart is likely timing out waiting on Spanning Tree. In Cisco switches set Spanning-Tree portfast (or Spanning-Tree portfast trunk) to configure the switch to begin passing traffic immediately.
that is a very curious problem. in core nothing has changed on the eda between 081 and 085. the changes all relate to the web interface and the scriptbuilder!
but something has to have changed. did you reimport the pxe boot files? what happens if you select a NIC? does it contiue or does it keep asking you for settings? are there any more people out there with this problem?
very well noted. our best practice states that on cisco, portfast always needs to be enabled. the installer is just too fast for the negotiation of some cisco--NIC combinations on normal installations this is wat our trunkport configurations looks like:
switchport trunk encapsulation dot1q
switchport trunk native vlan 21
switchport mode trunk
speed auto 1000
no cdp enable
spanning-tree portfast trunk
Yes, I can confirm that this problem exists, when the installer asks which NIC to choose. I was working with a fresh EDA server and reimported the PXE files for that EDA.
ah, found the problem. if you go to /var/www/eda and edit 'index.php' you need to change line 94 to:
append ip=dhcp initrd=initrd.img kssendmac ksdevice=$ksdevice load_ramdisk=1 ramdisk_size=10240 ks=http://";
somehow the 'ksdevice=$ksdevice' bit has gone missing. i'll update the downloadable image so people who dont know how to edit the php file can just re-download the appliance image.
I was wondering if you were going to add an update system to the app. It's kinda a pain to re-setup the app every new version or change. Since at least right now it seems to be mainly php changes would you release just those changes as something we could download, and apply/replace on the appliance?
a systemwide filelevel update requires a very different kind of update mechanism. between the apt-get updates it's going to take twice the effort to keep track of changes that dont come from the repositories. i doubt i'll find the time to do that between the development work to keep track of that 😛
there's a backup/restore option for the configuration though. it shouldn't be too much more trouble to unzip the appliance on the host and restoring the backup i think?
Great appliance! One question though...
We are a global company with several ESX datacenters around the world. ach one requires it's own install script, which varies slightly from the default due to hardware and networks. My question is, can this appliance create multiple scripts, with being able to assign a specific script to a host? I see a heading under the new host definition for "ks" script, but I don't see any way to have multiple scripts stored and selctable for each host that is entered.
Is this possible with this device? Yet?
@kcvman: I don't think that is possible yet, but one workaround you could do is running multiple EDA servers with separate zones, each with different ks script defined. It's a bit of a pain and a configuration hassle to do that way though.
i've had that request before, even to support multiple esx versions simultaniously (for training labs). but by design that's not what it's intended to do. perhaps in the post-v1.0
as for your setup, since EDA needs to be in the same segment as the ESX service console, it won't install over WAN links. so you'll need one in every site anyway. that makes it very easy to create a site-specific script too
usually we put the EDA on the first ESX server we deploy which we install from our workstation. just fastscp it over, vmkfstools -i, new VM, existing disk, done.
an import/export functionality would be handy indeed, or an "active script"and save standby / passive would be handy for us consultants.
If you find this information useful, please award points for "correct" or "helpful".
How would you go about performing an attended installation without the kickstart script, using the appliance?
if you define the servernames in the appliance you could start it at the bootprompt with 'esx01 ks=dummy'. it will produce an error saying something about an illegal kickstart path but it will start the normal attended installation wizard after that.
I would think this wouldn't be too hard, excluding the networking issues. My thought would be to have multiple kickstarts, each profile as it is needs to have the kickstart specified - it's just a matter of having the kickstart file/location change depending on which script you want - this would need to specified per server before hand. - For the moment I would think it'd be easier to have multiple EDA servers. The other thought is that I believe you can specify in a kickstart script to look at the SN/ and other information in the BIOS of the server. I'll look around again to see if I can find it, but at least in HP, you could specifiy a location field, and then you could have scripts for certain models/locations.
there's a backup/restore function that allows for as many scripts as you want. apart from a fileselection dialog it basically does exactly what describe. is there anything you would need next to that?
i actually have the scripts ready that when the post is done, captures the current config connected to the installing mac address. i didnt include them because typing 'restore' as a boot option is actually more typing than 'esx1' which in the end does exactly the same thing. and a restore would restore the old configuration while you may have altered the setup to included new insights or scripted things even further. it's probably a matter of choice. if you guys want the option i'll include it in the next release.