Setup

The basic setup here is to start 6-nodes where there are 3 master nodes and each master node has a slave node. This example setup uses Redis 3.2.8

Note: Replace all IP-addresses in redis.conf of each 700x directory for the key bind with non-loopback addresses.

Startup the cluster

./start_all.sh start

Create the replicas

Note: Replace the IP with a non-loopback address.

./redis-trib.rb create --replicas 1 192.168.1.117:7005 192.168.1.117:7001 192.168.1.117:7002 192.168.1.117:7003 192.168.1.117:7004 192.168.1.117:7000

Shutdown the cluster

./start_all.sh stop

Gotchas:

What to do when you see "Unable to connect to 127.0.0.1:7000"

Apparently, Redis disallows loopback and what you need to do is to replace the loopback address with the actual IP.

Reset the cluster when you run into "ERR Slot XXXX is already busy"

Make sure you stop servers first via start_all.sh stop

What you need to do is to remove all the configuration files that was indicated in the key cluster-config-file for each instance.

Restart the servers using start_all.sh start

The re-run the re-creation of the replicas, for example:

./redis-trib.rb create --replicas 1 192.168.1.117:7005 192.168.1.117:7001 192.168.1.117:7002 192.168.1.117:7003 192.168.1.117:7004 192.168.1.117:7000
>>> Creating cluster
>>> Performing hash slots allocation on 6 nodes...
Using 3 masters:
192.168.1.117:7005
192.168.1.117:7001
192.168.1.117:7002
Adding replica 192.168.1.117:7003 to 192.168.1.117:7005
Adding replica 192.168.1.117:7004 to 192.168.1.117:7001
Adding replica 192.168.1.117:7000 to 192.168.1.117:7002
M: 58e41945d5602e3ba347a992f4b7c3b13856dcbb 192.168.1.117:7005
   slots:0-5460 (5461 slots) master
M: 268253132bb74c5c50c42e97db6383a45c83f0bd 192.168.1.117:7001
   slots:5461-10922 (5462 slots) master
M: 4bc7a0254f3f446c0f742f1342404dcaba2161b8 192.168.1.117:7002
   slots:10923-16383 (5461 slots) master
S: e55772a802be0f2b3ae97c921ddca4b3f1e1871a 192.168.1.117:7003
   replicates 58e41945d5602e3ba347a992f4b7c3b13856dcbb
S: 215ec99d70af1225d65a06cb9e81ac3252948ee3 192.168.1.117:7004
   replicates 268253132bb74c5c50c42e97db6383a45c83f0bd
S: aaddedf219cfc0ef71f10d79f5c11fc4f6397c85 192.168.1.117:7000
   replicates 4bc7a0254f3f446c0f742f1342404dcaba2161b8
Can I set the above configuration? (type 'yes' to accept): yes
>>> Nodes configuration updated
>>> Assign a different config epoch to each node
54720:M 15 Jun 11:13:34.402 # configEpoch set to 1 via CLUSTER SET-CONFIG-EPOCH
54716:M 15 Jun 11:13:34.403 # configEpoch set to 2 via CLUSTER SET-CONFIG-EPOCH
54717:M 15 Jun 11:13:34.403 # configEpoch set to 3 via CLUSTER SET-CONFIG-EPOCH
54718:M 15 Jun 11:13:34.403 # configEpoch set to 4 via CLUSTER SET-CONFIG-EPOCH
54719:M 15 Jun 11:13:34.403 # configEpoch set to 5 via CLUSTER SET-CONFIG-EPOCH
54715:M 15 Jun 11:13:34.404 # configEpoch set to 6 via CLUSTER SET-CONFIG-EPOCH
>>> Sending CLUSTER MEET messages to join the cluster
54720:M 15 Jun 11:13:34.434 # IP address for this node updated to 192.168.1.117
54715:M 15 Jun 11:13:34.436 # IP address for this node updated to 192.168.1.117
....

raymondtay / Redos