redis-mmm 0.1.0 → 0.1.1

data/README.md

@@ -0,0 +1,82 @@
+redis-mmm - redis multi master replication manager
+==================================================
+
+This project is a port of a lot of ideas of mysql-mmm to redis. That also explains the name as an hommage to the mysql-mmm project
+
+Multi-Master?
+-------------
+
+No. Not really.
+
+
+What then?
+----------
+
+redis-mmm aims to provide automatic (and, to some extent, manual) failover functionality in case the redis master crashes or anything else happens.
+This functionality does not exist in redis itself.
+
+The whole idea behind that is having a virtual IP address to which all redis clients connect to reach the master. This IP is then migrated to the
+current master machine, so that the client will always reach the master. There is a slave IP address, too, so you could even balance the reads off of
+the master!
+
+All work is done by a separate monitoring process (which, for obvious reasons, should reside on a completely different host).
+
+
+How to make it work
+-------------------
+
+Before you start, you need to make sure you have the following prerequisites:
+
+* a network where you can dynamically change IPs (should be a private network)
+* 2 free IP addresses
+* you are NOT on Amazon EC2, since you can't add / remove IPs there (that's a lesson I had to learn)
+* a separate monitoring host on your network
+* two redis servers on different machines (I'll call them db1 and db2)
+* redis is listening on 0.0.0.0
+* the monitoring host can access both redis instances (I'll call it mon)
+* a redis-mmm user on all hosts
+* redis-mmm@mon can ssh to redis-mmm@db1 and redis-mmm@db2 without entering a password
+* redis-mmm@db{1,2} can use sudo to do the following commands
+  * /sbin/ip addr {show,add,del}
+  * /usr/bin/arping
+
+
+Now, go to the monitoring host, install redis-mmm (`gem install redis-mmm`) and configure /etc/redis-mmm.conf
+
+Example configuration:
+
+    master_ip = 192.168.10.70
+    cluster_interface = eth1
+
+    [db1]
+    address = 192.168.10.1
+    port = 6379
+    ssh_user = redis-mmm
+    ssh_port = 22
+
+    [db2]
+    address = 192.168.10.2
+    port = 6379
+    ssh_user = redis-mmm
+    ssh_port = 22
+
+
+To start the monitoring, start `redis_mmm mon`. It's left as an excercise for the reader to use upstart, god or bluepill to keep that process running.
+
+To view the cluster's status, call `redis_mmm status`
+
+To view info about a node, use `redis_mmm info <node>`
+
+To show help: `redis_mmm help`
+
+
+
+Warning
+=======
+
+This is an experimental piece of software not yet ready for prime time. I can't guarantee
+consistency for now, since there's no way to make a redis server read only and some redis clients
+seem to ignore crashed servers
+
+Nevertheless, play and try, fork and update and give feedback!
+

data/bin/redis_mmm

@@ -65,7 +65,7 @@ class RedisMMM < Thor
           :master_link_status => info["master_link_status"],
           :master_last_io_seconds_ago => info["master_last_io_seconds_ago"]
         }
-      rescue Errno::ECONNREFUSED
+      rescue Errno::ECONNREFUSED, Errno::EHOSTUNREACH
         {
           :online => false
         }
@@ -73,12 +73,11 @@ class RedisMMM < Thor
     end
 
     def add_ip(ip, interface)
-      $log.debug ssh("ip addr show #{interface} | grep #{ip} || sudo ip addr add #{ip} dev #{interface}")
-
+      $log.debug ssh("ip addr show #{interface} | grep #{ip} || (sudo ip addr add #{ip}/32 dev #{interface}; sudo arping -c 5 -U -s #{ip} -I #{interface} #{ip})")
     end
 
     def remove_ip(ip, interface)
-      $log.debug ssh("ip addr show #{interface} | grep #{ip} && sudo ip addr del #{ip} dev #{interface}")
+      $log.debug ssh("ip addr show #{interface} | grep #{ip} && sudo ip addr del #{ip}/32 dev #{interface}")
     end
 
     def slaveof(host)
@@ -107,6 +106,8 @@ class RedisMMM < Thor
       cmd = "/usr/bin/ssh -p #{ssh_port} #{ssh_user}@#{address} \"#{command}\""
       $log.debug "Executing #{cmd}"
       `#{cmd}`
+    rescue Errno::ECONNREFUSED, Errno::EHOSTUNREACH
+      'unreachable'
     end
 
   end
@@ -293,10 +294,10 @@ class RedisMMM < Thor
       return
     end
 
-    $log.info "Changing master from #{@current_master.name} to #{host.name}"
+    $log.info "Changing master from #{@current_master ? @current_master.name : "NONE"} to #{host.name}"
 
     # remove master ip from old master
-    remove_master_ip_from(@current_master)
+    remove_master_ip_from(@current_master) if @current_master
     # add master ip to new master
     add_master_ip_to(host)
     # let new master stop replication

metadata

@@ -1,13 +1,13 @@
 --- !ruby/object:Gem::Specification 
 name: redis-mmm
 version: !ruby/object:Gem::Version 
-  hash: 27
+  hash: 25
   prerelease: 
   segments: 
   - 0
   - 1
-  - 0
-  version: 0.1.0
+  - 1
+  version: 0.1.1
 platform: ruby
 authors: 
 - Michael Siebert
@@ -15,7 +15,7 @@ autorequire:
 bindir: bin
 cert_chain: []
 
-date: 2011-11-17 00:00:00 +01:00
+date: 2011-12-22 00:00:00 +01:00
 default_executable: 
 dependencies: 
 - !ruby/object:Gem::Dependency 
@@ -70,7 +70,7 @@ extensions: []
 extra_rdoc_files: []
 
 files: 
-- README
+- README.md
 - bin/redis_mmm
 has_rdoc: true
 homepage: http://github.com/siebertm/redis-mmm

data/README

@@ -1,32 +0,0 @@
-== Moving parts ==
-
-* redis_mmm_mond
-  checks connection to all hosts
-  checks replication state
-  assigns roles
-  emits SLAVEOF commands
-  connects via ssh to hosts and assigns the role ip's
-
-* redis_mmm_ctl
-  CLI
-  display cluster status
-  ability to change active master
-
-
-Example configuration:
-
-    master_ip = 192.168.10.70/24
-    cluster_interface = eth1
-
-    [db1]
-    address = 127.0.0.1
-    port = 6380
-    ssh_user = admin
-    ssh_port = 22277
-
-    [db2]
-    address = 127.0.0.1
-    port = 6381
-    ssh_user = admin
-    ssh_port = 22277
-