Display full list of DNS zones in VNet panel

[ravicious]

Demo recording

dns-zones.mov

At the moment, Connect only shows proxy hostnames of root clusters in the VNet panel. This PR makes it so that it uses the same logic as the VNet admin process to list DNS zones of clusters the user is currently logged into.

The first half of commits is related to lib/teleterm, the rest is related to the UI code.

UX choices

The list is fetched each time the user opens the VNet panel.

To establish the DNS zones of a cluster, VNet needs to ping it and then fetch vnet_config. The response built from those two sources is cached for 5 minutes. For root clusters, VNet also needs to fetch the list of leaf clusters – this response isn't cached but could be in the future. So if someone repeatedly opens and closes the panel, we only ever fetch the list of leaf clusters.

The VNet admin process has its own ClusterConfigCache. Every 10 seconds, the admin process updates the DNS configuration based on the state of TELEPORT_HOME_DIR, so essentially it asks the cache every 10 seconds for a config. This also means that it is able to configure VNet for new clusters or deconfigure it for stale clusters as the user logs in and out of clusters.

For 95% of users, the DNS zones will always remain the same. They only ever use a single cluster. For the other 5% of users, the list should get updated whenever they log in or out of clusters. Hence why we fetch the list each time the panel gets opened, instead of maintaining a second layer of caching that reacts to cluster changes in the Electron app.

The fact that the list won't change for most users is also why on subsequent openings of the panel we optimistically show stale values. On top of that, the status indicator behind updating DNS zones does not kick in until a certain delay. All of this to make sure that those 95% of users have least annoying UI possible – they should observe no changes on subsequent openings of the panel.

Perf choices

In Connect, ClusterConfigCache is shared between the VNet service and NetworkStack. Since NetworkStack requests a config on each new connection and Connect is going to request config for each cluster whenever the panel is opened, I figured it'd be best to share the cache.

I'm not entirely sure if it's going to cause any issues, e.g. Connect wants to show the DNS zones in the UI, so it asks for a config, which causes NetworkStack to process a TCP connection for which it also needs a config. The reason it works is because NetworkStack short-circuits when a connection is made to the cluster proxy hostname itself and does not fetch a config in that case.

---

changelog: VNet panel in Teleport Connect now lists custom DNS zones and DNS zones from leaf clusters

[ravicious]

The PR includes both Go and JS changes, but I don't expect both of you to review both parts. I guess @nklaassen will be more interested in lib/teleterm changes and @gzdunek in the UI changes.

[nklaassen]

Suggested change

	// Just like the admin process, it skips root and leaf clusters for which DNS couldn't be fetched
	// Just like the admin process, it skips root and leaf clusters for which the vnet_config couldn't be fetched

[public-teleport-github-review-bot]

@ravicious See the table below for backport results.

Branch	Result
branch/v16	Failed