komodorio / helm-dashboard Goto Github PK

Recognise & show ArgoCD-originating charts/objects, those helm ls does not show. They still contain annotations of being created by helm.
Maybe have a dedicated section of "Provisional Charts"
Argo's application CRD may be of help: https://github.com/argoproj/argo-cd/blob/master/manifests/crds/application-crd.yaml

Right now it is dirty with experiments and ugly things like obscure parameter names.
We need to make it more REST-like structure.
Designing an OpenAPI document looks like the right first step.

I miss the option to deploy it to K8s directly. I guess packing the helm binary and the plugin into a container and wrapping it into a deployment with a proper kubeconfig would work (?)

By default, the web server is only available locally. You can change that by specifying HD_BIND environment variable to the desired value. For example, 0.0.0.0 would bind to all IPv4 addresses or [::0] would be all IPv6 addresses.

so,how i use this HD_BIND to change localhost:8080,i cant find any answer

Bubble up the most interesting items: Deployments, StatefulSet, DaemonSet
Others remain unsorted.

Maybe bubble up the unhealthy.

Inside repository charts list, it can be long. Filtering would be nice
Should be a quick UI filter without re-querying the backend.

Hello, I noticed that Datadog RUM is being used for application telemetry.

I didn't see any disclosures about that (including what's being tracked, how it's being used, how that data is secured). Not in the console (during installation and runtime), webUI, or in the README.md.

Additionally; IMO, it would be good to make this telemetry require an explicit opt-in. Some people, if they're comfortable, might be glad to provide telemetry. Others might conclude that your plugin is spyware. Some orgs may have strong objections to the values of the name and namespace query parameters being exposed as they may contain sensitive data.

I assume that you have good intentions and just want to ship performant and reliable software. 👍

Even though I don't see any calls to startSessionReplayRecording(), It is good that you have defaultPrivacyLevel as mask. It still would be courteous to see an explicit set of sessionReplaySampleRate to 0 as a failback safety mechanism.

off-topic: I do like the plugin! It's extremely useful.

Thanks!

https://docs.brew.sh/Formula-Cookbook

Probably it takes the wrong field to find release's source. It should look at "chart" field, parse it to take only the name part and not version, and search it in the repositories. The release "name" is a wrong field to take.

I've overlooked that on the PR review here: https://github.com/komodorio/helm-dashboard/blob/main/pkg/dashboard/static/list-view.js#L32

Right now it's only possible to configure it via env variable and less experienced people are confused (#69, #72).
Let's introduce a CLI flag for that. Long-only flag --bind would be fine.
Don't forget to mention it in the README.md

Find out how to make navigation to remember the hash history and reload page on clicking Back.

It's already there in the hash string, just need to arrange a place for it in UI.

The function should be safe to run on any cluster name. Currently, there are errors like
clusterSplit.at is not a function. (In 'clusterSplit.at(-1)', 'clusterSplit.at' is undefined on some cluster names.

Use kubeconfig and appropriate env vars to figure out configuration, but don't require kubectl binary to be present, because Helm doesn't. Maybe even reuse API client that Helm has inside its codebase.

Hi Komodor IO

I'm getting:

Is it possible for the plugin here to open the default browser/figure out what it is and open it?

Thank you.

when I try to use helm dashboard --help on my system, it seems to print the version and exit.

$ helm dashboard --help
INFO[0000] Helm Dashboard by Komodor, version 0.2.3 (549cdd9bfbdf32009f8dbbc240c59c86c2e430d7 @ 2022-10-26T14:27:14Z)

I am on Ubuntu Jammy Jellyfish, using helm 3.10.1

While displaying the helm chart in the helm dashboard long chart goes out of the box.
Here is a screenshot of the issue: https://imgur.com/a/Is8j0Ry

I'm using supervisor to manage the startup process

The configuration file is as follows

[program:helm-dashboard]
command=helm dashboard -b -p 8888
priority=1                    
numprocs=1                    
autostart=true                
autorestart=true              
startretries=10               
exitcodes=0                   
stopsignal=KILL               
stopwaitsecs=10               
redirect_stderr=true          

stdout_logfile_maxbytes = 1024MB
stdout_logfile_backups  = 10
stdout_logfile          = /var/log/supervisor/helm-dashboard.log

The process is not started successfully

$ supervisorctl status
helm-dashboard                   FATAL     Exited too quickly (process log may have details)

However, the startup log reports the following error

tail /var/log/supervisor/helm-dashboard.log 
Error: unknown command "dashboard" for "helm"
Run 'helm --help' for usage.
Error: unknown command "dashboard" for "helm"
Run 'helm --help' for usage.
Error: unknown command "dashboard" for "helm"
Run 'helm --help' for usage.
Error: unknown command "dashboard" for "helm"
Run 'helm --help' for usage.
Error: unknown command "dashboard" for "helm"
Run 'helm --help' for usage.

The log error appears to be incorrectly identifying the helm dashboard command
How do I correctly identify the helm dashboard command as if it were executed from the command line

By default, the web server is only available locally. You can change that by specifying HD_BIND environment variable to the desired value. For example, 0.0.0.0 would bind to all IPv4 addresses or [::0] would be all IPv6 addresses.

How does HD_BIND specify listener 0.0.0.0 when started with a command

In the Install dialog, offer existing namespaces in drop-down. Keep the ability to enter arbitrary namespace.

DisruptionAllowed - why do we see this status? Is it red or green? Maybe we should look at different field for status.

Relates to #32 with regard of our general approach of telling the health status of a resource.

I was trying to access a k8s cluster where I have access to only a single namespace, but when I tried to run helm dashboard command from CLI on that k8s cluster it gave me an error as it's trying to scan all the namespaces which i don't have access to

WARN[0018] helm ls --all --all-namespaces --output json --time-format 2006-01-02T15:04:05Z07:00 --kube-context xxxxx
WARN[0018] STDERR:
Error: list: failed to list: secrets is forbidden: User "zzzz" cannot list resource "secrets" in API group "" at the cluster scope

I would suggest having 2 features here which would be very useful

1. A parameter to set the default namespace which can be used on startup and other operations
2. Ability to switch namespace from the UI (if the user has access to multiple namespaces on the k8s cluster)

We already have API that serves current tool version, need just to display it somewhere.
Most likely, in the top bar.

It would be great if dev server can detect changes in static and refresh the UI

helm dashboard --port 9999 INFO[0000] Helm Dashboard by Komodor, version 0.2.3 (549cdd9bfbdf32009f8dbbc240c59c86c2e430d7 @ 2022-10-26T14:27:14Z)

When it’s running you get more output and it tells you something along the lines of, “Running on localhost:9999”, but I cannot access the dashboard itself, see below.

Any help would be appreciated

Right now, the resources from dependent charts are not included into views, and generally no additional information is visible about it.

To avoid broken releases

Hello,
when I try to install or update the plugin I encounter the following error:

/bin/sh: 0: Illegal option -
Error: plugin install hook for "dashboard" exited with error

Using GNU bash, version 5.1.16(1)-release (x86_64-pc-linux-gnu) on WSL2 (Ubuntu 22.04.1 LTS)

If I try to use helm dashboard anyway I get

Error: fork/exec /home/vmariette/.local/share/helm/plugins/helm-dashboard.git/bin/helm-dashboard: no such file or directory

To be able to open releases in a new browser tab, intuitively.

Installed charts in list should be correct links, and full row should be a link, not just header text.

Top menu items Installed/Repository should be correct links.

Revision links on the left should be correct links.

For example, under different name/into different namespaces.

That would require changes around InstalledName, maybe with list of installed.

Look at its structure critically.
Maybe split it into several data classes.
Add documentation to each method.

On the Resources tab, display all container images with their tags, to ease understanding of image versions.

Current idea is the button that would open a flyout with list of tags and their respective resources.

On the backend, we can gather that info while getting status information for resources. The image information is not supposed to change for particular revision, so it's safe to cache.

we have helm that version is pre-release and only have pre-release version, while dashboard default search stable version, so I can not find my helm chart in the Repository tab.

Can helm-dashboard`s Repository tab support to display helm that only have pre-release versions ?

just add --devel flag in helm search repo command can get the pre-release version.

ref: https://helm.sh/docs/helm/helm_search_repo/

# Search for release versions matching the keyword "nginx", including pre-release versions
$ helm search repo nginx --devel

edit

func (d *DataLayer) ChartRepoVersions(chartName string) (res []*RepoChartElement, err error) {
	search := "/" + chartName + "\v"
	if strings.Contains(chartName, "/") {
		search = "\v" + chartName + "\v"
	}

	cmd := []string{"search", "repo", "--devel", "--regexp", search, "--versions", "--output", "json"}
	out, err := d.runCommandHelm(cmd...)
	if err != nil {
		return nil, err
	}

	err = json.Unmarshal([]byte(out), &res)
	if err != nil {
		return nil, err
	}
	return res, nil
}

func (d *DataLayer) ChartRepoCharts(repoName string) (res []*RepoChartElement, err error) {
	cmd := []string{"search", "repo", "--devel", "--regexp", "\v" + repoName + "/", "--output", "json"}
	out, err := d.runCommandHelm(cmd...)
	if err != nil {
		return nil, err
	}

	err = json.Unmarshal([]byte(out), &res)
	if err != nil {
		return nil, err
	}

	ins, err := d.ListInstalled()
	if err != nil {
		return nil, err
	}

	enrichRepoChartsWithInstalled(res, ins)

	return res, nil
}

or give a args to control this feature.

Current dialog is too noisy and slow.

Rework it into a wizard-like full-page experience. Experiment to see if this is a better approach. Rework value editing into smart diff calculator.

loki example - DaemonSet and StatefulSet better status display

Right now we don't show good status for those.

Right now it is hardcoded as "default". We have an optional default namespace coming from context information. We should use that (if present) as namespace when installing new chart.
In case there is forced namespace set, use it instead.

SERVLET: org.eclipse.jetty.servlet.ServletHandler$Default404Servlet-5f8edcc5

After doing a particular scan the report is incomplete and cant be scrolled. A download button to download the complete scan or maybe a scrollable button will be helpful.

Right now it's too much of callbacks.
Something like async/await-based client class would be much more appropriate.

Instead, change the "Check for new version" into a message to add the missing repository.

This was mentioned in comments of #23

Query GH API, compare with current version, offer to upgrade

Error received from tabs: "Installed" and "Repository"

helm-dashboard version v0.2.0
helm version "v3.2.1"

When managing a helm chart via Flux a HelmRelease custom resource is created with a given storageNamespace (foo), but could have a different targetNamespace (bar) which determines where the chart is installed (ref the spec here). The net result is that helm secrets, etc will be stored in the foo namespace, but bar is the "release namespace" where the chart is installed.

The logic handling of helm "lookups" by the dashboard seems to be problematic in this scenario. See the below example:

> helm ls -n foo
NAME                         	NAMESPACE     	REVISION	UPDATED                                	STATUS  	CHART                     	APP VERSION
bizbang                      	bar       	1       	2022-10-24 14:16:07.803938 -0600 MDT   	deployed	bizbang-1.45.0

From this helm command you can see that the namespace scope of the helm ls command is foo (my Flux storageNamespace), but the namespace listed here is bar (my Flux targetNamespace).

When I try to click on this in the dashboard I receive the error "Failed to get chart details Error: release: not found". Looking at my terminal I can see it is using the wrong namespace in the lookup:

WARN[1087] Failed command: [helm history bizbang --namespace bar --output json --max 18 --kube-context <removed>] 
WARN[1087] STDERR:
Error: release: not found

Using the foo namespace the helm history command would make this work.

It would be great to see this scenario supported/handled, although I understand if its seen as an unsupported edge case.

We need to address a case when user has local Helm chart and wants to enjoy the preview of manifests.
Probably it should be analogous to chart repositories.

Maybe filter by name. Think of using release name + chart name + namespace as fields to search for.

Definitely a filter by namespace. Probably as radio buttons, with option "All"

If there is forced namespace, and that namespace is different from kubectl config default, pre-select it. Relates to #51

The idea: attach to .Log() calls and communicate the lines of progress back to UI.
Example long installing is airflow chart.

In the list of installed charts, attempt to improve display by issuing additional call to query for icon and description.
This would require introducing a new API call, since we don't want to delay the display of the list. This needs to match release to repository, which we already able to do.
Then, helm show chart komodorio/k8s-watcher to get the icon URL and description.
Make sure the UI is not "jumping" because of variable description length.

I experienced a complete roadblock to using this system. I work against a multi-tenant kubernetes managed by our cloud team, and the tokens that I used to authenticate kubectl contexts do not have RBAC policies allowing them to list anything at the cluster level. I am using:

• helm v3.10.1
• dashboard version 0.2.3
• Ubunutu Jammy Jellyfish

When I launch helm dashboard with a --namespace=$mynamespace argument, it still wants to list all namespaces, and fails. Once it fails there is no way to choose a specific namespace. It just sits there, completely unusable.

helm dashboard --namespace=londo003
INFO[0000] Helm Dashboard by Komodor, version 0.2.3 (549cdd9bfbdf32009f8dbbc240c59c86c2e430d7 @ 2022-10-26T14:27:14Z) 
WARN[0000] Failed command: [checkov --version]          
WARN[0000] Failed command: [trivy --version]            
INFO[0000] Opening web UI: http://localhost:8080        
Gtk-Message: 09:57:07.641: Failed to load module "canberra-gtk-module"
Gtk-Message: 09:57:07.643: Failed to load module "canberra-gtk-module"
WARN[0000] Failed command: [helm ls --all --all-namespaces --output json --time-format 2006-01-02T15:04:05Z07:00 --kube-context londo003-ocp] 
WARN[0000] STDERR:
Error: list: failed to list: secrets is forbidden: User "system:serviceaccount:londo003:helm-deployer" cannot list secrets at the cluster scope: no RBAC policy matched 
WARN[0126] Failed command: [helm ls --all --all-namespaces --output json --time-format 2006-01-02T15:04:05Z07:00 --kube-context londo003-ocp] 
WARN[0126] STDERR:
Error: list: failed to list: secrets is forbidden: User "system:serviceaccount:londo003:helm-deployer" cannot list secrets at the cluster scope: no RBAC policy matched 
WARN[0131] Failed command: [helm ls --all --all-namespaces --output json --time-format 2006-01-02T15:04:05Z07:00 --kube-context londo003-ocp] 
WARN[0131] STDERR:
Error: list: failed to list: secrets is forbidden: User "system:serviceaccount:londo003:helm-deployer" cannot list secrets at the cluster scope: no RBAC policy matched 
WARN[0146] Failed command: [helm ls --all --all-namespaces --output json --time-format 2006-01-02T15:04:05Z07:00 --kube-context londo003-ocp] 
WARN[0146] STDERR:
Error: list: failed to list: secrets is forbidden: User "system:serviceaccount:londo003:helm-deployer" cannot list secrets at the cluster scope: no RBAC policy matched 
^C