Resolve slow key retrieval when some GPG key servers are not reachable

[h3xcat]

At the moment, if the host environment cannot reach first key server over port 11371, it takes 1 minute to retrieve individual keys from the other listed key servers. For the Python feature, the installation script retrieves 13 keys, so the total time to retrieve all the keys is 13 minutes. The delay could be further extended if subsequent servers also cannot be reached.

The reason it takes 1 minute rather than the 15 second default timeout in dirmngr is that I believe dirmngr internally tries the connection to the same keyserver 4 times before it moves on to a different server. Essentially, this multiplies the timeout delay by 4.

Instead of relying on the dirmngr timeout mechanism, the change implements a check for keyserver reachability using curl before the list is passed on to dirmngr.

If no keyservers are identified, the script will intentionally fail. If some keyservers are not reachable, the script will log this information, enabling users to more easily identify and resolve network issues in their environments.

Related issues:
#686
#337
#323

Affected features:

• git-lfs
• git
• github-cli
• kubectl-helm-minikube
• python
• ruby
• terraform

I ran tests for all the features locally; most succeeded without issues. The only test that failed was autoPullDisabled from git-lfs, but it also fails with unmodified code, so I suspect something else in my environment is causing the tests to fail, not specific to my changes.

[h3xcat]

@microsoft-github-policy-service agree

[samruddhikhandale]

thanks for contributing, definitely a helpful improvement.

[samruddhikhandale]

Looks good, thanks! Let's merge this on Monday to ensure we accidentally don't break anything.

[h3xcat]

@samruddhikhandale
Could you merge the pull request when you are able to? If there are any concerns, please let me know.