Azure: Sample apps: Terraform tests

[DrisDary]

Changes

1. All Terraform tests used the same resource group name "local-rg", causing conflicts when tests ran in parallel on CI. Changes include assigning unique prefixes to each test sample.
2. Sqlcmd with ODBC Driver 18 rejects LocalStack's self-signed SSL certificate by default therefore added -C flag (trust server certificate) to all sqlcmd commands.
3. Not all terraform commands work with python 3.13 so had to revert to python 3.12.
4. Due to disk usage limit error, function app deployment timeout errors, I now run tests in 4 shards instead of 2 allowing more resources free resources per test.

To Do:

1. For one of the test i.e. function-app-managed-identity sometimes we get the error:

Status Code: 500, Details: {"error": true, "exception": "URLError", "message": ""}

This is a results of the function _check_function_status in our function deployer code which has a timeout set to 1 second for the urlopen. This needs to be increased.

2. The runners we use to run these large sample tests are relatively weak and can result in timeouts and errors such as API rate limit error. In the future we need to either upgrade these runners or move to a repo that allows for powerful runners.

[paolosalvatori]

@DrisDary thanks for building this workflow. You don't need a real review, you know. But here are a few comments for you:

• I see that two of the four jobs still fail. Is this due to the fact that runners do not have enough resources in terms of CPU and memory?
• I see that you modified main.tf files to add a resource_group_name = "${var.prefix}-rg" local variable, and then you modified terraform.tfvars files to add a unique default name for the prefix variable, one for each sample.
• Do you delete the current resource group before provisioning the next Terraform sample? That could slow down a bit the execution, but reduce the amount of resource required, as you remove the containers used by previous samples.

[paolosalvatori]

Approved. Just reply my comments. ☝️

[DrisDary]

@DrisDary thanks for building this workflow. You don't need a real review, you know. But here are a few comments for you:

• I see that two of the four jobs still fail. Is this due to the fact that runners do not have enough resources in terms of CPU and memory?
• I see that you modified main.tf files to add a resource_group_name = "${var.prefix}-rg" local variable, and then you modified terraform.tfvars files to add a unique default name for the prefix variable, one for each sample.
• Do you delete the current resource group before provisioning the next Terraform sample? That could slow down a bit the execution, but reduce the amount of resource required, as you remove the containers used by previous samples.

• So one is due to a localstack timeout which is fixed by deleting the github cache and rerunning. The other error is an API rate limit error which i believe means the sum of the api calls made accross all running tasks is limited and exceeding leads to failure so in our case its across 4, again rerunning only failed runs should fix this since there will be less tests to run after you rerun failures. Our localstack-pro repo is more advanced and has no such limitations in terms of api rate limit errors or weak runners. [I will rerun the failed tests to see what happens]

• yes I did change names and then also made sure it was consistent through out the terraform files this is because sometimes same names results in conflicts and hence failures.

• yes deletion occurs inbetween all tests to save disk usage and is insignificant in terms of cpu usage or even time. The next test doesnt start until the last test is completely deleted so it should not cause an resource usage errors.

[DrisDary]

@paolosalvatori it appears it worked after rerunning just as expected.

[paolosalvatori]

@DrisDary if you see 429 errors, these are due to throttling (to many requests). The only solution in this case is retrying.