kubectl --kubeconfig cannot merge multiple files #468
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
KnVerey
force-pushed
the
revert_config_flag
branch
2 times, most recently
from
7385fe8
to
75853c4
Compare
KnVerey
commented
Apr 18, 2019
| @@ -28,13 +28,14 @@ def test_build_client_from_multiple_config_files | |||
| Kubeclient::Client.any_instance.stubs(:discover) | |||
| # Set KUBECONFIG to include multiple config files | |||
| dummy_config = File.join(__dir__, '../../fixtures/kube-config/dummy_config.yml') | |||
| kubeconfig = "#{ENV['KUBECONFIG']}:#{dummy_config}" | |||
There was a problem hiding this comment.
This wasn't providing great multi-config coverage in our CI, which doesn't set the env var.
| @@ -27,7 +26,6 @@ def run(*args, log_failure: nil, use_context: true, use_namespace: true, output: | |||
| output_is_sensitive = @output_is_sensitive_default if output_is_sensitive.nil? | |||
|
|
|||
| args = args.unshift("kubectl") | |||
| args.push("--kubeconfig=#{@kubeconfig}") | |||
There was a problem hiding this comment.
KnVerey
commented
Apr 18, 2019
| def build_kubeclient(api_version:, context:, endpoint_path: nil) | ||
| # Find a context defined in kube conf files that matches the input context by name | ||
| configs = config_files.map { |f| KubeConfig.read(f) } | ||
| config = configs.find { |c| c.contexts.include?(context) } | ||
| raise TaskConfigurationError, "No kubeconfig files found in #{@kubeconfig_files}" if kubeclient_configs.empty? |
There was a problem hiding this comment.
This is new. As CI pointed out, without it, if you try to build a kubeclient without validating the config files, and one of them is missing, we would 💥 with an Errno::ENOENT. My thinking with the way I addressed this is:
- We don't need to be re-reading the config files every time we build a new client, hence the new memoization above
- If for whatever reason the code did not run the validation before building a client, and we have both a valid config containing the desired context and a missing file, we should go ahead and do what the user wanted. If we can't do what the user wanted, we should throw our own more helpful error, not
Errno::ENOENT.
timothysmith0609
approved these changes
Apr 22, 2019
KnVerey
force-pushed
the
revert_config_flag
branch
from
966ffe7
to
4b14166
Compare
KnVerey
force-pushed
the
revert_config_flag
branch
from
255432c
to
a67a4ea
Compare
ghost
added
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What are you trying to accomplish with this PR?
Revert only the problematic part of #428, i.e. https://github.com/Shopify/kubernetes-deploy/pull/428/files#diff-f5478b2ac59d4babc30470442f3a8656. I still like the Class refactor, but
--kubeconfigcannot merge multiple files, so if we want to use it we'll need to be more clever about what we pass through. Ironically the test we removed would have caught this regression. I'm sorry I did not catch that. I added it back verbatim (after confirming it does reproduce the bug on master).Fixes #463
How is this accomplished?
Making the
kubectlclass go back to depending on the env var.As a bonus, make it proactively validate this dependency.after CI showed this broke default location support, I removed it to keep this PR the smallest viable fix.I think there is a nicer way to deal with this, but I wasn't able to put together a proposal as quickly as I wanted to. I think we should merge something really simple like this so we have time to discuss the right refactor to do.
What could go wrong?
If somebody consuming this project as a gem switched away from the environment variable, they'll need to switch back (it is still effectively mandatory for CLI usage).
@Shopify/cloudx