Multiple instances on same host?

[poempelfox]

I've been wondering whether it's possible to run multiple instances of mirrorbits (meaning: for different projects, with a different set of mirrors) on the same host.
As you can use multiple config files and the config does allow you to set RedisDB which then seems to select a different DB, this should probably work?
However, the instances then still detect each other on startup: 2018/01/28 11:04:07.140 CET -> Node mirrorbits-21868 joined the cluster
As far as I understood the code, this is because it does some broadcast communication through the database, and that just uses the static string "HELLO" as a marker.
My attempt to fix this was to add the RedisDB-number to the string like that:

--- a/daemon/cluster.go
+++ b/daemon/cluster.go
@@ -11,6 +11,7 @@ import (
        "sync"
        "time"

+       . "github.com/etix/mirrorbits/config"
        "github.com/etix/mirrorbits/database"
        "github.com/etix/mirrorbits/mirrors"
        "github.com/etix/mirrorbits/utils"
@@ -94,6 +95,7 @@ func (c *cluster) Stop() {
 func (c *cluster) clusterLoop() {
        clusterChan := make(chan string, 10)
        announceTicker := time.NewTicker(1 * time.Second)
+       clusterAnnWithId := clusterAnnounce+fmt.Sprintf("%d", GetConfig().RedisDB)

        c.refreshNodeList(c.nodeID, c.nodeID)
        c.redis.Pubsub.SubscribeEvent(database.CLUSTER, clusterChan)
@@ -106,18 +108,18 @@ func (c *cluster) clusterLoop() {
                case <-announceTicker.C:
                        c.announce()
                case data := <-clusterChan:
-                       if !strings.HasPrefix(data, clusterAnnounce+" ") {
+                       if !strings.HasPrefix(data, clusterAnnWithId+" ") {
                                // Garbage
                                continue
                        }
-                       c.refreshNodeList(data[len(clusterAnnounce)+1:], c.nodeID)
+                       c.refreshNodeList(data[len(clusterAnnWithId)+1:], c.nodeID)
                }
        }
 }

 func (c *cluster) announce() {
        r := c.redis.Get()
-       database.Publish(r, database.CLUSTER, fmt.Sprintf("%s %s", clusterAnnounce, c.nodeID))
+       database.Publish(r, database.CLUSTER, fmt.Sprintf("%s%d %s", clusterAnnounce, GetConfig().RedisDB, c.nodeID))
        r.Close()
 }

That seems to work, the instances no longer see each other. Is that really all that's needed, or am I overlooking something here that will cause problems down the road?

[etix]

I'm astonished to read that there was no scoping in pubsub... so you're right we need to include the database index in the channel name.

Your patch is pretty good but while you're at it please make a function to generate the pubsub channel name instead of using Sprintf in two different locations.

[etix]

Pull request merged.