delete VM not working via TRE API about azuretre

[marrobi]

Looks like a permission denied issue. When you move scripts on a linux file system you may need to grant execute permission using chmod +x ./delete_vm_extensions.sh prior to building the bundle.

Let us know how you get on.

[Danny-Cooke-cCK]

within VScode the files do have the correct permissions.

[Danny-Cooke-cCK]

when copying the windows template from azuretre to our deployment. can i copy just that folder? or do i have to copy the whole structure including the guacamole server templates too ?

[marrobi]

Ok, can you check within the bundle, find the image ID of the cloukubed-azure-windowsvm image using docker images

Then run:

vscode ➜ /workspaces/marrobi-azure-tre (main) $ docker run -it 84c482413687  bash
nonroot@921195ae47d2:/cnab/app$ ls -ltr
total 56
-rwxrwxr-x 1 root root  584 Nov  2  2022 reset_password.sh
-rwxrwxr-x 1 root root 1122 Nov  2  2022 delete_vm_extensions.sh
-rw-rw-r-- 1 root root 1336 Feb  2 22:22 template_schema.json
-rwxrwxr-x 1 root root 1635 Apr 29 14:18 parameters.json
-rw-rw---- 1 root root 9958 Jun 23 09:38 porter.yaml
-rwxrwx--- 1 root root  117 Jun 23 09:38 run
drwxrwxr-x 1 root root 4096 Jun 23 09:38 runtimes
drwxrwx--- 1 root root 4096 Jun 23 09:38 mixins
drwxrwxr-x 1 root root 4096 Jun 23 09:38 terraform

Check the permissions, and try running the delete_vm_extensions.sh script (it will fail, but should run)

[Danny-Cooke-cCK]

i dont appear to have one running locally?
however i do have docker and containers running

sudo docker images | grep cloukubed-azure-windowsvm
returns nothing ?

[marrobi]

The command I used runs one in interactive mode.

Have you missed a d from cloudkubed?

[Danny-Cooke-cCK]

that was just a typo sorry.
sudo docker images | grep cloud
returns nothing.

so my vscode is opened in dev container but i did all the deployment in git pipeline. Have a missed a step, setting up the docker images locally?

[marrobi]

Ah, ok, if just using a pipeline you are unlikely to be able to debug locally.

To build and register locally use make user_resource_bundle BUNDLE=cloudkubed-azure-windowsvm WORKSPACE_SERVICE=guacamole

Registration might fail depending on local config, but the image should build and then you will see it in docker images.

[Danny-Cooke-cCK]

i tried the above command and a few variations of it but failing but it is most definitely there at that location.

/bin/bash: line 4: cd: /home/vscode/AzureTRE//templates/workspace_services/guacamole/user_resources/cloudkubed-azure-windowsvm: No such file or directory
make[1]: *** [/home/vscode/AzureTRE/Makefile:183: bundle-build] Error 1
make[1]: Leaving directory '/workspaces/tre11d'
make: *** [/home/vscode/AzureTRE/Makefile:284: user_resource_bundle] Error 2

inside my porter file is this

schemaVersion: 1.0.0
name: tre-service-guacamole-windowsvm-ck
version: 0.7.4
description: "An Azure TRE User Resource Template for Guacamole (Windows 10)"
dockerfile: Dockerfile.tmpl
registry: azuretre

[marrobi]

Does this folder exist?

/home/vscode/AzureTRE//templates/workspace_services/guacamole/user_resources/cloudkubed-azure-windowsvm

[Danny-Cooke-cCK]

yes it does. Its most definitely there. with the right permissions etc

[Danny-Cooke-cCK]

also here is the original AzureTRE templates i copied from

[marrobi]

Can you try just running porter build in the directory? Thanks.

[marrobi]

Actually, worth checking, is the repo checked out to Windows or Linux file system?

[Danny-Cooke-cCK]

porter build has completed successfully.

this is checked out to Windows with WSL running underneath, docker running too etc

subsequent command however still failed
make user_resource_bundle BUNDLE=cloudkubed-azure-windowsvm WORKSPACE_SERVICE=guacamole

/bin/bash: line 4: cd: /home/vscode/AzureTRE//templates/workspace_services/guacamole/user_resources/cloudkubed-azure-windowsvm: No such file or directory
make[1]: *** [/home/vscode/AzureTRE/Makefile:183: bundle-build] Error 1
make[1]: Leaving directory '/workspaces/tre11d'
make: *** [/home/vscode/AzureTRE/Makefile:284: user_resource_bundle] Error 2

[marrobi]

Ok, I reckon that could be the issue. There is a PR to add something to the docs for #3551

At the end of the porter build do you see:

 => => writing image sha256:2d1f9128e899dbe3ceeebb2e7268614a351a99b754331a58163a442bcc6bbf14                                                                                                                                                                                                   0.0s
 => => naming to docker.io/azuretre/tre-service-guacamole-linuxvm:porter-6932c599d5e3c317413ce4547b0a62f1                                                                                                                                                                                      

Then run:

docker run -it <docker.io/azuretre/tre-service-guacamole-linuxvm:porter-6932c599d5e3c317413ce4547b0a62f1> bash

With your image name as per porter output and check the permissions.

We typically checkout inside WSL, rather than on Windows and think that could be the cause.

[Danny-Cooke-cCK]

thanks ok so a slight difference between windows and linux.

here was the output

and i beleive this is the image i need to connect to now ?

once i've got through all this i will put a PR in with documentation updates etc

[marrobi]

You are missing bash of the end of the command to get a prompt inside the container where you can check the permissions.

[Danny-Cooke-cCK]

indeed i did! good spot. All permissions look fine inside

[marrobi]

Very strange. My next steps would be to try installing/uninstalling locally using make bundle-install and bundle-uninstall. Is a bit tricky.

You are building the images in the pipeline right? So they will be on a Linux filesystem there. I'm a bit confused, might have to check if anybody else has any ideas.

[Danny-Cooke-cCK]

i forgot to say as well.
i can run the command fine once logged in
./delete_vm_extensions.sh <resource_group> tfstate f548bd1f-8cf7-41b0-a41d-1909ce5657ce

./exec-runtime uninstall did literally nothing but hung for 30 mins

[Danny-Cooke-cCK]

Also over the weekend i've tested in a second environment and also with Linux and im experiencing the same behaviour

[Danny-Cooke-cCK]

Would it potentially behave differently if we put our templates in the same location as the default templates in azuretre? Also i assume other people have custom templates with custom VM's in place without an issue ?

[marrobi]

Let me check how other customers are using the custom VM templates.

[marrobi]

So to build user resource templates in the deployment repo you can use:

DIR=./path/to/my/templates BUNDLE_TYPE=user_resource WORKSPACE_SERVICE_NAME=tre-service-guacamole make bundle-build bundle-publish bundle-register 

As for the delete issue, still not clear why that is happening, will talk with the wider team.

[Danny-Cooke-cCK]

Thanks @marrobi . Yes we are building like that but obviously via the pipeline rather than manual.

[Danny-Cooke-cCK]

@marrobi have you guys had chance to look at this ?

I was wondering, could the issue be related to using images from an internal gallery instead of images from the market place ?

[marrobi]

@Danny-Cooke-cCK did you get this resolved?