Installing a plugin requires restart of main octant process.

[n3wscott]

I installed the sample plugin and it did not show up until I restarted.

[imjasonh]

+1, This also makes plugin development a bit more painful, since I have to rebuild my plugin and restart Octant to see changes.

[jdolitsky]

Hi, I ran into the same issue. Needing to restart octant to check plugin changes during development etc.

I put together this script to autobuild the plugin and restart octant on source changes: https://gist.github.com/jdolitsky/34b67113afe515e096064a4d01fd4dd6

[jdolitsky]

Would it be useful to add some flag like --watch-plugins that periodically re-checks/re-loads the plugin directory?

[bryanl]

Octant should be able to restart a plugin or plugins. Or perhaps you kill the plugin and Octant could could restart registered plugins. We will draft upma solution and tag the issue.

[imjasonh]

Just heard of this script to help with plugin development, which looks promising: https://gist.github.com/jdolitsky/34b67113afe515e096064a4d01fd4dd6

I haven't tried it yet, but I plan to when I have time to play with plugin development again.

[jdolitsky]

Octant could restart registered plugins

That would work great. I just tried to kill the plugin process manually, and Octant falls into a bad state, w/ this in the logs over and over:

err in navigation: grpc client response: rpc error: code = Unavailable desc = all SubConns are in TransientFailure, latest connection error: connection error: desc = "transport: error while dialing: dial unix /var/folders/51/0bdn93qj1s98v438m1y357lw0000gn/T/plugin228550586: connect: connection refused"2019-11-16T23:37:51.773-0600	ERROR	api/navigation_manager.go:97	load namespaces	{"poller-name": "navigation", "err": "grpc client response: rpc error: code = Unavailable desc = all SubConns are in TransientFailure, latest connection error: connection error: desc = \"transport: error while dialing: dial unix /var/folders/51/0bdn93qj1s98v438m1y357lw0000gn/T/plugin228550586: connect: connection refused\""}
github.com/vmware-tanzu/octant/internal/api.(*NavigationManager).runUpdate.func1
	/go/src/github.com/vmware-tanzu/octant/internal/api/navigation_manager.go:97
github.com/vmware-tanzu/octant/internal/api.(*job).run.func1
	/go/src/github.com/vmware-tanzu/octant/internal/api/poller.go:111

The webapp remains available, but the lefthand menu then disappears:

Problem resolved by restarting octant

[wwitzel3]

My initial thought for doing something like this is to extend the existing process launcher to have a bit more management to it.

• add fsnotify/fswatch to the plugin folder
• compare the filename that changed to our filename list in our process management
• kill/restart the plugin with a matching filename that has changed

Related reading:

• Handling plugin crashes hashicorp/go-plugin#31
• Resumable plugins rebased vmware-tanzu/velero#710