Custom installed UCS has reachability issue on AWS EC2

Hi everyone,

I am recently trying to set up a UCS on AWS for some internal use. The Hong Kong region is preferred but unluckily the official image is not available there. Therefore, I tried to install from scratch by myself. The details are as follows:

  1. I created a VM with VirtualBox using the official bootable ISO image.
  2. I started the VM and installed UCS according to the instructions in 2. Creating a UCS appliance/cloud image — Univention Corporate Server - Extended installation documentation.
  3. I further configured the VM by running the commands in 2. Creating a UCS appliance/cloud image — Univention Corporate Server - Extended installation documentation.
  4. I exported the VM with the OVF2.0 format and then imported it into AWS as an AMI via the command aws ec2 import-image.
  5. I launched an EC2 with the imported AMI.
    Consequently, I obtained a running EC2 on AWS but I cannot SSH into it (error is connection timeout). As I tested that SSH is working with some other linux image for the same VPC setting, I believe the issue is caused by some misconfiguration in my VM, e.g., network adaptor, network interface name, etc.

Any help is greatly appreciated!

I have done OVF from a VM a couple of times for UCS.
and something is always broken some place…, if not immediately noticeable… some where down the line.
Usually related to the partitions…

but definitely super flaky as regards the networking… it really hates for the network names to be changed.
it is not stored just in the flat linux files, but squirreled away in the UCS variables and stored as FIXED names

like “ens192”, but sometimes the image might come up as “ens132” or some variation during initial setup.

so then it cannot find it…

the way i might tackle this… is mount the disk images as a separate drive to another linux installation.
and check the names allocated by amazon for their network interfaces

then add these into the linux images as secondary network NICS when you build the initial images.
so … dont just config 1 interface when building the image locally, connfig ,some with DHCP & some fixed
but leave the base one (nic 0) for ucs to play about.

It was a long running issue with linux on the names generated & given to nics & the order they came up, during booting…
even today… multiple linux options don’t have this finalised…

Thank you for the detailed response. Unfortunately enabling 2 network interfaces on the VM doesn’t solve my issue. Probably not going to try anymore. I am wondering if Univention will officially copy the images to other regions.