homeport/havener

/ˈheɪvənə/ - Think of it as a swiss army knife for Kubernetes tasks

Havener /ˈheɪvənə/

Table of Contents

• Introduction
• Ok, tell me more
• How do I get started
• Contributing
• License

Introducing Havener

If you use a Kubernetes cluster, chances are very high that you use kubectl a lot. These are fine tools and allow you to do everything you need to do, but there are use cases where you end up with a very long kubectl command in your terminal. This is why we created havener to introduce a convenience wrapper around kubectl. Think of it as a swiss army knife for Kubernetes tasks. Possible use cases are for example executing a command on multiple pods at the same time, or retrieving usage details.

Ok, tell me more

To see a detail list of all havener commands, please refer to the command documentation.

Like kubectl, havener relies on the Kubernetes configuration that can be set via the KUBECONFIG environment variable. It can also be provided with the --kubeconfig flag, which takes the path to the YAML file (for example $HOME/.kube/config).

Notable Use Cases

Usage overview

Quickly get a live overview of the current cluster usage details, for example Load, CPU, and Memory of the cluster nodes.

Pods in namespace overview

Watch pods in multiple namespaces with added colors to help identify the respective state.

Havener Commands

• havener events - Show Kubernetes cluster events
• havener logs - Retrieve log files from all pods
• havener node-exec - Execute command on Kubernetes node
• havener pod-exec - Execute command on Kubernetes pod
• havener top - Shows CPU and Memory usage
• havener watch - Watch status of all pods in all namespaces

How do I get started

There are different ways to get havener. You are free to pick the one that makes the most sense for your use case.

• On macOS systems, a Homebrew tap is available to install havener:

  brew install homeport/tap/havener

• Use a convenience script to download the latest release to install it in a suitable location on your local machine:

  curl -sL https://raw.githubusercontent.com/homeport/havener/main/scripts/download-latest.sh | bash

• Of course, you can also build it from source code (only for development):

  go install github.com/homeport/havener/cmd/havener

Contributing

We are happy to have other people contributing to the project. If you decide to do that, here's how to:

• get Go (havener requires Go version 1.20 or greater)
• fork the project
• create a new branch
• make your changes
• open a PR.

Git commit messages should be meaningful and follow the rules nicely written down by Chris Beams:

The seven rules of a great Git commit message

1. Separate subject from body with a blank line
2. Limit the subject line to 50 characters
3. Capitalize the subject line
4. Do not end the subject line with a period
5. Use the imperative mood in the subject line
6. Wrap the body at 72 characters
7. Use the body to explain what and why vs. how

Running test cases and binaries generation

There are multiple make targets, but running all does everything you want in one call.

make all

Test it with Linux on your macOS system

The best way to test is to use Docker to spin up a container:

docker run \
  --interactive \
  --tty \
  --rm \
  --volume $GOPATH/src/github.com/homeport/havener:/go/src/github.com/homeport/havener \
  --workdir /go/src/github.com/homeport/havener \
  golang:latest /bin/bash

Package dependencies (Go modules)

The Go module setup can be frustrating, if you have to update Kubernetes API libraries. In general, using go get with a specific version based on a tag is known to work, for example go get k8s.io/[email protected]. In case you run into difficulties, please do not hesitate to reach out to us.

License

Licensed under MIT License

havener's Issues

Report panic: runtime error: index out of range

Executing the command upgrade with the --config flag causes the following error.
According to @HeavyWombat, the bug is probably in /internal/hvnr/diffs.go:75.

Error Message:

runtime error: index out of range

Stacktrace:

runtime/debug.Stack(0xc0000e3690, 0x2140fbe, 0x11)
	/usr/local/go/src/runtime/debug/stack.go:24  0x9d
github.com/homeport/havener/internal/cmd.exitWithErrorAndIssue(0x216b748, 0x31, 0x232dd40, 0x2edf600)
	/Users/lukas.mueller/Projects/havener/internal/cmd/common.go:106  0x287
github.com/homeport/havener/internal/cmd.Execute.func1()
	/Users/lukas.mueller/Projects/havener/internal/cmd/root.go:58  0x171
panic(0x1fba5c0, 0x2edf600)
	/usr/local/go/src/runtime/panic.go:522  0x1b5
github.com/homeport/havener/internal/hvnr.ShowHelmReleaseDiff(0xc000150150, 0x7, 0xc0001501b0, 0xe, 0xc0003ec3c0, 0x3c, 0x40, 0x0, 0xc0003ae000, 0x32756d0)
	/Users/lukas.mueller/Projects/havener/internal/hvnr/diffs.go:75  0x769
github.com/homeport/havener/internal/cmd.UpgradeViaHavenerConfig(0x7ffeefbffad7, 0x1b, 0xe, 0x7ffeefbffad7)
	/Users/lukas.mueller/Projects/havener/internal/cmd/upgrade.go:112  0x4fe
github.com/homeport/havener/internal/cmd.glob..func9(0x2eeade0, 0xc0003cc0e0, 0x0, 0x2, 0x0, 0x0)
	/Users/lukas.mueller/Projects/havener/internal/cmd/upgrade.go:56  0x64
github.com/spf13/cobra.(*Command).execute(0x2eeade0, 0xc0003cc040, 0x2, 0x2, 0x2eeade0, 0xc0003cc040)
	/Users/lukas.mueller/Projects/golang/pkg/mod/github.com/spf13/[email protected]/command.go:826  0x465
github.com/spf13/cobra.(*Command).ExecuteC(0x2eea8e0, 0x21cc5a0, 0x21cddd8, 0x109326f)
	/Users/lukas.mueller/Projects/golang/pkg/mod/github.com/spf13/[email protected]/command.go:914  0x2fc
github.com/spf13/cobra.(*Command).Execute(...)
	/Users/lukas.mueller/Projects/golang/pkg/mod/github.com/spf13/[email protected]/command.go:864
github.com/homeport/havener/internal/cmd.Execute()
	/Users/lukas.mueller/Projects/havener/internal/cmd/root.go:66  0x60
main.main()
	/Users/lukas.mueller/Projects/havener/cmd/havener/main.go:42  0xcd

Need a better message format when deployment failed

When deployment failed, havener returned this message

binaries/havener-darwin-amd64  deploy --config examples/tomcat-deployment.yml
⠙ Creating Helm Release for tomcat-releaseMessage: Unable to create and initialize the tunnel, Error: could not find tiller

It contained two sentences.

Maybew we can split by : or other format:

⠙ Creating Helm Release for tomcat-release: Message: Unable to create and initialize the tunnel, Error: could not find tiller

Add travis CI

Install tiller if not available

On a fresh cluster (e.g. minikube), there is not set-up for helm available. This means there is no tiller pod.

Proposal: Perform something like helm init as part of havener for the deploy command.

Add operator to retrieve a secret

Add an operator similar to the (( shell ... )) operator to retrieve a secret from a given namespace and secret.

Is it possible to have a new parameter for top command so that just show one time

Right now, the havener top command will keep reporting top result.

I think in the future, it can be automated so that an automation script run one time to save the result for debugging or just for show for us.

So keep reporting may not work for that. Can we add a new parameter for example havener top -n 1 to just show the result 1 or x times?

Move ExitWithError out of havener package

The ExitWithError function is designed to provide an explanation to the end-user. It should be moved into the internal package. If possible, avoid using os.Exit() and experiment with runtime.Goexit() so that potential defer functions are honoured.

Replace client-go with kubernetes-sigs/controller-runtime

Describe the bug
cr

To Reproduce
Steps to reproduce the behavior:

1. Go to '...'
2. Click on '....'
3. Scroll down to '....'
4. See error

Expected behavior
A clear and concise description of what you expected to happen.

Screenshots
If applicable, add screenshots to help explain your problem.

Additional context
Add any other context about the problem here.

Modify helm setup in favor of host helm binary

Havener could use under the hood the helm binary of the host where it runs, this will decrease the complexity of managing different dependencies that are compatible between them(e.g. kube, go, helm)

@HeavyWombat correct me if im wrong?

Add `--no-cross-compile` flag to build script

Add --no-cross-compile flag to build script so that only a binary for the current platform is created. This flag should then be used in the Dockerfiles to avoid compiling macOS binaries if only the Linux one is required.

havener version should output helm and k8s API version

Since havener bundles other APIs, let the version command output not just the havener version, but also the version of the packaged dependencies like helm and k8s API.

Add alternative root shell option

Investigate whether the following alternative would be feasible:

kubectl run r00t --restart=Never -ti --rm --image lol --overrides '{"spec":{"hostPID": true, "containers":[{"name":"1","image":"alpine","command":["nsenter","--mount=/proc/1/ns/mnt","--","/bin/bash"],"stdin": true,"tty":true,"securityContext":{"privileged":true}}]}}'

Add new blueprint functionality

Describe the bug
A clear and concise description of what the bug is.

To Reproduce
Steps to reproduce the behavior:

1. Go to '...'
2. Click on '....'
3. Scroll down to '....'
4. See error

Expected behavior
A clear and concise description of what you expected to happen.

Screenshots
If applicable, add screenshots to help explain your problem.

Additional context
Add any other context about the problem here.

Change type of chart version to string

The mongoDB example chart can't be deployed because the env operator is a String but the field is unmarshaled as Integer before the operator is resolved. Because version numbers are often more complex than usual Integers, it makes sense to change the field type to String.

Havener purge must be more resilient

Failed to delete a Helm Release, which was in state FAILED:

secondsMessage: Error deploying chart, Error: rpc error: code = Unknown desc = release uaa failed: Service "uaa-uaa-public" is invalid: spec.externalIPs[0]: Invalid value: "a.root-servers.net.\nb.root-servers.net.\nc.root-servers.net.\nd.root-servers.net.\ne.root-servers.net.\nf.root-servers.net.\ng.root-servers.net.\nh.root-servers.net.\ni.root-servers.net.\nj.root-servers.net.\nk.root-servers.net.\nl.root-servers.net.\nm.root-servers.net.": must be a valid IP address, (e.g. 10.9.8.7)

Shell operator processing won't work with windows

Update: In theory, we can support havener configuration files with (( shell ... )) operator that can do things in Windows. However, we have to discuss if it makes sense to put time into this since it is currently not a priority to have havener fully functional on Windows systems.

Chart location should support a file endpoint

On any config.yml, the https://github.com/homeport/havener/blob/develop/examples/minikube-config.yml#L9 should support downloading releases from a github file, e.g. tar or zip , and specifically select a file inside it.

Rewrite Verbose functions with vargs

To avoid something like verbose(fmt.Sprintf("%d", num)), let's try to refactor the code so that verbose() itself accepts vargs and a format string: verbose("num: %d", num).

@ansrev Could be something for you :)

Show output of deployment steps in real-time (no buffering)

Describe the bug
In case the pre-deployment step takes very long, one does not understand what is currently happening.

To Reproduce
Use a havener configuration with a pre-deployment step, which takes 120 s and prints out a status every second.

Expected behavior
Live output of the pre-deployment.

Screenshots
n/a

Additional context
n/a

Remove unused TODOs strings from cmd information

Add logs retriever

Investigate terminal size issue

The terminal size overriding rule for garden containers does not seem to be working.

Make docker image for node-exec configurable

Currently, the docker image used for getting a shell on a node relies on a docker image available on DockerHub. Make it configurable.

Code clean-up/refactor for user versus backend logic

In general, we need to move to a design where end-user specific code like spinners, error message printing, verbose messages, or confirmations are solely in the internal packages. Everything else should be in the pkg/havener package.

Remove `chart_` prefix from configuration items

Rename chart_name to just name and so on to have a cleaner configuration.

Logs retrieval command must not download configuration files by default

Change logs retrieval command to only download log files by default. Add an optional flag to download configuration files that might contain secrets and credentials.

Rework naming and description

The whole repo name and description should align itself to what the cli actually do, which is more like a toolkit for operating a kubernetes cluster. It is also more appealing.

Add timeout to pod readiness probe in node-exec

Add a timeout error to tge readiness probe of the created pod in the 'node-exec' function to avoid long wait times if the pod is unable to start.

Add helm upgrade mode

Add pkg versioning

There are two major dependencies with delicate versioning requirements: client-go and helm. It could be that we have to introduce a versioning of the havener package itself to reflect that different versions of that are required to match with code with specific Kubernetes API packages, or helm.

So far we have not hit an issue so far, but it would be nice to discuss this issue and potential problems.

Bring back terminal cursor if canceled during spinner is running

Bring back terminal cursor if canceled during spinner is running. Run havener to show a spinner. Ctrl-C out of it and the cursor is not reinstated.

Cannot install helm chart from URL

Add distributed shell option to node-exec/pod-exec

Add an option to run a command on multiple nodes or multiple pods at the same time.

Spinner elapsed time should jump to minutes

Change spinner elapsed time to display minutes instead of 234 s.

Enhance logs cmd to support custom directories

Describe the bug
Improve the logs cmd, to allow a user to specify an specific path inside the pod from which to get the logs

To Reproduce
DOES NOT APPLY

Expected behavior
User can specify the logs directory

Screenshots
If applicable, add screenshots to help explain your problem.

Additional context
Add any other context about the problem here.

Align stucture of pod-exec and node-exec

Make sure that both pod-exec and node-exec feel and behave the same, that means similar named flags for the same purpose, or same assumptions in case of missing arguments (e.g. no command, default to /bin/bash).

Add helm binary into the travis file

Dependabot can't resolve your Go dependency files

Dependabot can't resolve your Go dependency files.

As a result, Dependabot couldn't update your dependencies.

The error Dependabot encountered was:

build github.com/homeport/havener: cannot load github.com/homeport/dyff: cannot find module providing package github.com/homeport/dyff

If you think the above is an error on Dependabot's side please don't hesitate to get in touch - we'll do whatever we can to fix it.

You can mention @dependabot in the comments below to contact the Dependabot team.

Introduce logging feature, e.g. debug, info

Add flag to logs retrieval to compress downloaded logs

Add a flag to the logs retrieval command to compress the logs into a tarball.

Improve the top command from the content and look wise perspective

Describe the bug
A clear and concise description of what the bug is.

To Reproduce
Steps to reproduce the behavior:

1. Go to '...'
2. Click on '....'
3. Scroll down to '....'
4. See error

Expected behavior
A clear and concise description of what you expected to happen.

Screenshots
If applicable, add screenshots to help explain your problem.

Additional context
Add any other context about the problem here.

Add shell variables support to config

Support setting and using shell variables in before and after hooks.

Spike: Check whether havener package can provide a verbose message stream

Rather than printing the verbose messages immediately, check whether we can have an implementation where the havener package provides a message stream that can be consumed by the internal package so that the end-user code can decide when and how to show messages, or send to a log file.

Encapsulate parts from deploy.go, upgrade.go and exec.go

Especially deploy.go and upgrade.go are very similar and have a lot of redundant code. Together with exec.go which also features config reading and parsing, it make sence to encapsulate parts of the code to remove redundant/duplicated code.

Add filter flags for the event cmd

This will help to filter per namespace or resource.

Add command to do Pod exec

Similar to node-exec, introduce pod-exec to have an easy way of getting a shell in a pod. Ideally, you do not provide any namespace or pod name and havener will give you a choice of pods to exec into.

Update readme with latest new features

1. The --config file deserves a detail explanation in the README that covers the following features:

• pre steps
• post steps
• retrieval of secrets via (( secret .. ))
• usage of env variables via (( env ..))

Spike: Check how complex it gets a diff before upgrade

Check if dyff reports can be used to show a nice BOSH-like diff before doing an upgrade.

Havener spinner (progress indicator) UI glitch in Concourse

The spinner does not look ok running Havener in Concourse.

Buffer output in certs before printing it

Right now, certs subcommand will print the details in the moment when they are processed. Especially when the verbose mode is enabled, you end up with output and verbose output being mixed.

Proposal: Buffer the output in something like var buf bytes.Buffer and print the verbose message whenever it is required. Once all data is collected, print the content of the buffer as one block.

@ansrev That could be something for you to pick up.