redhat-developer / vscode-openshift-tools Goto Github PK

This is must fix, because current odo release only can operate on active project/app/component and that means to build tree vscode extension has to set active/project/app/component multiple times. It works only for simple hierarchy where is only one project, one app and multiple components. As soon as hierarchy grows to several project with several apps each there would be several sub-trees building in parallel and that is when tree is getting messed up.

Examples:

• odo project list' => oc get project -o json'
• 'odo app list' => has no oc alternative, because aaplication names saved in ~/.kube/odo file to be used as labels and name suffixes for real OpenShift resources created by odo commands
• odo list => can be build based on DeploymentConfig resources labels
• odo url list =>can be build based on Route resources labels
• odo storage list => oc get pvc -o json

Even better way to use -o jsonpath
oc get project -o jsonpath='{range .items[*]}{.metadata.name}{"|"}{.metadata.annotations.openshift\.io/display-name}{ "|"}{.metadata.annotations.openshift\.io/description}{"\n"}{end}' --no-headers=true

"OpenShift" tree is not telling the user what cluster it is listing, one has to know what cluster was selected as default cluster or what cluster you logged in via "oc" on the cmd-line.
This is even more confusing since the "Clusters" view can get out-of-sync if you login manually via oc to a different server: it wont loose it's children. You have to manually refresh to discover it.

steps:

1. ASSERT: have at least 2 servers
2. EXEC: in "Clusters" view: select server1 and pick "Set as current cluster"
3. ASSERT: in "Clusters" view: selected server1 gets arrows and can be unfolded to show resources
4. EXEC: in Terminal: use "oc login" to login into server2
5. ASSERT: you're logged in into server2
6. EXEC: in "OpenShift" view: hit refresh

Result:
"OpenShift" view shows your project on server2, "Clusters" view still looks as if server1 was your current cluster.

Btw. Since "Cluters" view is still thinking that server1 is the current cluster, it wont offer "set as current cluster" in the context menu. Refreshing the "Clusters" view helps though.

redhat-developer/odo#413

OpenShift Explorer tree should use the same icons DevStudio has in OpenShift Explorer.

VSCode Version: 1.26.1

Currently there is no indication of the progress of the commands which are performed in vscode. Provide a progress item to indicate the same.

It would be great if the README would provide the steps on how to install this extension into VSCode. Currently a user that has no intimate knowledge on developing VSCode extensions has no clue how to achieve that.
Then a user that stumbles upon this repo would love to get some info on what this is and what it provides. Currently the README is just a stub.

Add command that runs `odo project delete' for OpenShift project.

oc CLI released only in zipped format. Download should support zipped downloads and unzip them after download is finished.

in package.json the name reads "vscode-openshift-tools" while display name is "OpenShift Connector".
That would make the extensions id "redhat.vscode-openshift-tools".

Make the id more in lines with the display name, and mostly shorter if possible.

@mohitsuman , @dgolovin , thoughts?

Test would verify extension commands can

• execute odo commands without errors
• parse odo commands output correctly

Test should watch odo nightly build location and run test for every new odo build.

Test should be executed on local OpenShift cluster.

Would be good to be able to see component's log without updating it everytime.

odo log -f

Current process assumes user knows what is going on and can select appropriate component type available in list generated by 'odo catalog list'.

Follow the same rules throughout all context menus in OpenShift explorer. Every menu should have at least three groups:

• new commands
• info commands
• watch/push commands if applicable
• delete command

View's Navigatio Menu:

Context Menu for a project item:

Context menu for Application item

Context menu for component node
**

**

There is only way to add new cluster to kubernetes view by using oc login https://clusterurl -u username in console. There should be command that would let do the same from vscode UI.

Since I work for quite some time with different cdk's etc my list of local clusters is huge:

I'd love to be able to select them all and remove them at once. Currently this isn't possible, I have to do that one by one :(

"Clusters" view, that's contributed by the microsoft kubernetes extension, does not work with the cdk nor openshift online.
In the context of our OpenShift tooling for VSCode it's purpose is limited to select the server that shall be used to display ODO resources. It thus causes a bad/buggy/flawed impression to the user:

a noob user that has a kube config files already would ask himself:

• "why is the cdk/openshift online server not listing anything?"
• "why can't I add my openshift online server?"
• "why can't I add my cdk server?"

I think that we should seriously consider replacing it by a much simpler view to select a server or fix the kube tooling.

Now OpenShift explorer has:

• Delete Component
• Delete Application
• Delete Project

in OpenShift Explorer context menus

All this actions should be named as Delete

• #671 Provide user names from k8s config when login to the cluster

API https://code.visualstudio.com/docs/extensionAPI/vscode-api#QuickInput
Example: https://github.com/Microsoft/vscode-extension-samples/tree/master/quickinput-sample
Screenshot:

'odo catalog list services'

Current implementation have to use three commands to set active object before executing object specific command. That approach obviously does not work for multiple threads because threads can override each other. This for example leads to wrong OpenShift Explorer tree content for clusters with more than one projects with multiple applications each. Using oc/kubectl can help solve the problem, but in this case vscode-extension should know how to distinguish odo created objects form other ones using labels/annotations.

Add MIT license and name for vscode extension.

odo acatalog command changed to support new entity 'service' in odo model. Now odo catalog can list:

• components as `odo catalog list components'
• services as `odo catalog list services'

This command would allow to create example components from repositories mentioned in s2i builder image description.
Should appear as new command in Command Plette and as a quickpick item when called from OpenShift Application Explorer.

If dc for components has url to repository it was created with, extension can clone it to workspace using git

The reason is hardcoded path to shell executable '/bin/bash'

About action should print out all available version information:

• odo version
• OpenShift version
• kubernetes version

the workflow using the "OpenShift (ODO)" view is to

1. select a current cluster - which you do in in the "Clusters" view and then
2. look at what this server holds in terms of "OpenShift (ODO)" resources.

In terms of mouse one thus currently has to skip "OpenShift" and go down to "Clusters" to select the server. Then back up to "OpenShift". The more "natural" flow would be to select the server in "Clusters" view, which is the 1st view that's shown, and then move your mouse further down to the "OpenShift" view to discover what this current cluster holds in terms of OpenShift resources.

Might be good place to show commands:

1. Push
2. Create OpenShift Component

"commands": [
      {
        "command": "openshift.project.deploy",
        "title": "New OpenShift Component",
        "category": "OpenShift"
      }
]

"menus": {
      "scm/sourceControl": [
        {
          "command": "openshift.project.deploy",
          "group": "navigation",
          "when": "scmProvider == git"
        }
      ]

OpenShift resource tree seems to only show apps that I created via ODO. Anything that I created outside of it is not shown.

To me the "OpenShift" resource tree is more like an "ODO" resource tree than an OpenShift resource tree. It thus should be labeled as such, otherwise it's just confusing.

Implementation should understand odo commands and provide the same stdout output for testing. It should retain created object in configuration file and generate 'odo list' output based on saved objects to behave like real 'odo' cli.

If new odo version is available, stdout contains additional strings

---
A newer version of odo (version: 0.0.11) is available.
Update using your package manager, or run
curl https://raw.githubusercontent.com/redhat-developer/odo/master/scripts/install.sh | sh
to update manually, or visit https://github.com/redhat-developer/odo/releases
---

they should be filtered out before parsing.

oc get route -o jsonpath='{range .items[?(.metadata.labels.app\.kubernetes\.io/component-name=="$component-name"})]}{.spec.host}{"\n"}{end}'
will return host for $component-name.
if
oc get route -o jsonpath='{range .items[?(.metadata.labels.app\.kubernetes\.io/component-name=="$component-name"})]}{.spec.tls.termination}{"\n"}{end}'
is not empty then protocol is https otherwise http.

then use https://www.npmjs.com/package/opn to open final url.

if route is not present then, info message should notify user about that and give two options:

• create urls and open
• cancel

If create url is selected then 'openshift.url.create' command should be executed and then 'openshift.component.open' command called again.

Current implementation for New Component Command supports only --local option.

I tried listing the project in the resource tree, but it does not show anything. In the terminal I noticed the following output:

Please login to your server: 

oc login https://mycluster.mydomain.com

• To me things are not obvious enough. One stumbles upon this output while refreshing repeatedly and noticing things moving in the output then.
• Furthermore one discovers that one has to install an additional tool, oc.
• Then things would be more user-friendly if "odo" would output the correct command (with correct url) right away instead of https://mycluster.mydomain.com

An idea that we discussed is to offer a "log in" cmd in the context menu.
Even more great would be if the tooling was doing that automatically and prompt the user to provide pw/token?

Current implementation just clone project into workspace subfolder. That does not allow vscode recognize it as project and it treats repository as reqular subfolder inside workspace.