We are installing Rhino on AWS EC2 instances running Windows Server 2025. Since service release 22, the following pop-up appears when starting Rhino for the first time (message A dynamic link library (DLL) initialization routine failed.). After killing the process and trying again, Rhino typically starts successfully. Sometimes, three tries are necessary. We are adapting our deployments scripts to work around this issue. Any idea what could be causing this?
Update: It seems we can’t work around this issue by adapting our deployment scripts. Retrying to start Rhino several times from a script did not resolve the issue. To resolve it, we have to manually log in to the EC2 instances using RDP. Therefore, this issue becomes more important because it means we can’t roll out SR 22 without manual user intervention.
Note: I also noticed the problem with service release candidate 23.
thanks for reporting,
this looks this issue:
RH-79832 RhinoCore.dll could not be loaded
I’m not sure why this is on Future right now, I’ll ask
Can you right click RhinoCore.dll in ProgramFiles\Rhino8\System, show Properties and go to Digital Signatures? then double click the digital signature, what does it show for you?
Hi @snabela. I wasn’t able to replicate this issue on my own VM running Windows Server 2025. I used the stock image from AWS, but I wasn’t sure if you were using a custom image or not. I updated my install script so that it simply tries to install Rhino on the machine (but it doesn’t try to install all of the other IIS/Rhino.Compute stuff). You can view the install script that I used here and if you wanted to try to run it on your own VM, you can do so by opening a powershell console (in Administrator mode) and running the following command:
iwr -useb https://raw.githubusercontent.com/mcneel/compute.rhino3d/8.x/script/production/module_install_rhino_only.ps1 -outfile install_rhino_only.ps1; .\install_rhino_only.ps1
This should install Rhino and when I ran the application, it started right away. You will need to know your core-hour-billing auth token when you run this script. Follow this for more information about the prerequisites.
We are using the stock Windows Server 2025 AMIs provided by AWS, an example:
Our install script looks almost the same as yours, besides we download the installer from a S3 bucket of our own, because we need to control which service release gets installed (up until SR 21 this problem didn’t happen).
Other than that, we are installing the .NET 8.0 Desktop Runtime prior to installing Rhino. Currently we are using version 8.0.16, but I see there is newer version 8.0.19.
Interestingly, once I log in to the EC2 instance using RDP and then try to start Rhino, the problem disappears on the second try.
Maybe Windows Defender is getting in our way here. It’s a miracle.
Yuck. This is going to be no fun to sort out.
We could do some bisecting by sending you a list of builds between when it last worked and when it first broke. Is this something you’re willing to do? If you’re willing, I can DM you a list of builds to try.
I assume @AndyPayne that you tested on the same image as @snabela?
Yes. I tested on the same image. If I could replicate the issue then I could bisect the issue but so far I haven’t seen the same error message.
Yes let’s do that, many thanks!
Note: Installing .NET 8.0 Desktop Runtime 8.0.19 instead of 8.0.16 did not help. Also, I tried a few of the recent versions of the Windows server AMI (they are being updated by AWS roughly once a month).
Are you running the installation script as a logged in user (using remote desktop), or from an EC2 metadata script?
As a logged in user.
Got it, that likely makes the difference. You could try to run the script headless (from EC2 metadata), then log in using RDP and start Rhino.
Ok. I’ll give it a try.
1 Like
I was able to run the script I posted above via the EC2 User Data input in the launch configuration (following this guide). I had to modify the script just a bit to hard-code the email, api key, and auth tokens (rather than prompting the user for those). I then logged into the VM and started Rhino and everything worked fine (no .dll error). Have you tried using the script I posted above as a test on your end?
I haven’t tried running exactly your script yet, but I trust that it works. However, I need to find out what’s causing the pop-up, so I will start bisecting using the builds provided by @brian.
Ah, one more thing that is different for our scripts: We create a non-admin user for running Rhino.
Hmm. We do the same with the standard bootstrap script that installs the full rhino.compute package. I didn’t necessarily test that full install (only the Rhino installation part of it). I’ll try to run the full test tomorrow and see if I can replicate the issue. So when you get this error, are you logged in as your non-admin user and launching Rhino that way?
Yep, our script follows this schema:
- fresh EC2 instance gets launched
- EC2 userdata is used to launch a PowerShell script that
- (among other things) installs Rhino
- creates a non-admin user and configures it for auto-logon
- configures another PowerShell script to be run on every logon (using the Run registry key)
- The second PowerShell script (now running without admin privileges) launches Rhino
1 Like
Did you get a chance to give it a try? Otherwise, if I can get access to the installers, we can do some bisecting.
I have been investigating this. So far, I haven’t found anything but will continue to dig in on our side to see if I can find anything.
1 Like