Issue #37: Added multi-cluster support for displaying cluster specifi…

…c information.

Issue #16: Added cluster-name in kubectl, notifier and ping messages from botkube.

config.yaml

@@ -125,5 +125,3 @@ settings:
   clustername: not-configured
   # Set false to disable kubectl commands execution
   allowkubectl: false
-  # Set true only respond to channel in config
-  checkchannel: false

design/multi-cluster.md

@@ -0,0 +1,51 @@
+# Multi-cluster Support
+
+#### Assumptions
+`@botkube` commands refer to all the commands in the slack bot which currently supports:
+- kubectl
+- notifier
+- ping
+
+
+### Summary
+Add Multi-cluster support for Botkube, where a single bot can monitor multiple clusters and respond to `@botkube` commands with cluster specific results.
+
+### Motivation
+Currently in multi-cluster scenario, a Slack bot authenticates all the clusters with a same authentication token. Thus running `@botkube` command returns response from all the configured clusters, irrespective of the slack channel or group. For `@botkube` command execution, we need a particular cluster specific output.
+
+### Design
+
+This design approach adds a flag `--cluster-name` to all `@botkube` commands. Use of that flag is optional in a cluster specific channel.
+
+Botkube `Notifier` commands are restricted to a dedicated channel for a cluster only and `--cluster-name` flag is ignored. 
+
+Botkube `ping` command with the `--cluster-name` flag returns `pong` response from the cluster specified in the flag, else you get response from all the clusters. `Ping` command without --cluster-name flag can be used to list all the configured clusters in the slack bot and identify you cluster's name among them.
+
+For `kubectl` commands in a dedicated channel to a cluster, if `--cluster-name` flag is used, it responds with the output for the cluster specified in flag, else it checks if the channel in the request matches the `config.Communications.Slack.Channel` and responds if true else ignores.
+
+For `kubectl` commands in a group, Direct message or channel not dedicated to any cluster, the `--cluster-name` flag is mandatory. The executor checks if the `--cluster-name` flag is present in the request. If yes, it gets the cluster's name from the flag and compares with `c.Settings.ClusterName` from the config file, if it matches then it responds with the required output to the slack bot and if it doesn't match, it ignores the request. And if the `--cluster-name` flag is absent for kubectl commands, it responds to the slack bot saying 'Please specify the cluster-name'. 
+
+For example -
+```sh
+@Botkube get pods --cluster-name={CLUSTER_NAME}
+```
+where,
+`CLUSTER_NAME` is the name of the cluster you want to query.
+
+To get the list of all clusters configured in the slack, you can run the following command in slack.
+
+```sh
+@Botkube ping
+```
+
+##### Workflow
+
+![Multi_Cluster_Design](workflow.png)
+
+
+### Drawbacks
+The `--cluster-name` flag is mandated for kubectl and notifier commands resulting additional overhead.
+
+### Alternatives
+We can add channel specific authentication token or completely dedicate a channel to a particular cluster which requires changes in the slack code.
+

helm/botkube/values.yaml

@@ -140,8 +140,6 @@ config:
     clustername: not-configured
     # Set false to disable kubectl commands execution
     allowkubectl: false
-    # Set true only respond to channel in config
-    checkchannel: false
 
 resources: {}
   # We usually recommend not to specify default resources and to leave this as a conscious

pkg/config/config.go

@@ -49,7 +49,6 @@ type Slack struct {
 type Settings struct {
 	ClusterName  string
 	AllowKubectl bool
-	CheckChannel bool
 }
 
 // New returns new Config

pkg/controller/controller.go

@@ -1,6 +1,7 @@
 package controller
 
 import (
+	"fmt"
 	"os"
 	"os/signal"
 	"strconv"
@@ -20,13 +21,13 @@ import (
 	"k8s.io/client-go/tools/cache"
 )
 
-var startTime time.Time
-
 const (
-	controllerStartMsg = "...and now my watch begins! :crossed_swords:"
-	controllerStopMsg  = "my watch has ended!"
+	controllerStartMsg = "...and now my watch begins for cluster '%s'! :crossed_swords:"
+	controllerStopMsg  = "my watch has ended for cluster '%s'!"
 )
 
+var startTime time.Time
+
 func findNamespace(ns string) string {
 	if ns == "all" {
 		return apiV1.NamespaceAll
@@ -39,7 +40,7 @@ func findNamespace(ns string) string {
 
 // RegisterInformers creates new informer controllers to watch k8s resources
 func RegisterInformers(c *config.Config) {
-	sendMessage(controllerStartMsg)
+	sendMessage(fmt.Sprintf(controllerStartMsg, c.Settings.ClusterName))
 	startTime = time.Now().Local()
 
 	// Get resync period
@@ -127,7 +128,7 @@ func RegisterInformers(c *config.Config) {
 	signal.Notify(sigterm, syscall.SIGTERM)
 	signal.Notify(sigterm, syscall.SIGINT)
 	<-sigterm
-	sendMessage(controllerStopMsg)
+	sendMessage(fmt.Sprintf(controllerStopMsg, c.Settings.ClusterName))
 }
 
 func registerEventHandlers(resourceType string, events []string) (handlerFns cache.ResourceEventHandlerFuncs) {