Differences between revisions 36 and 56 (spanning 20 versions)

Redis

Redis is an in memory data structure storage. It will be used to share PHP-sessions between the apache servers. To guarantee a robust deployment three redis servers will be configured.

192.168.1.61 redis (virtual IP-address)
192.168.1.62 redis01 (master)
192.168.1.63 redis02 (replica)
192.168.1.64 redis03 (replica)

Software

apt-get install redis-server redis-sentinel

Configuration

Redis Server

Change the contents of the redis configuration file in /etc/redis/redis.conf.

Add the IP-address for each of the servers. Each of the redis servers needs to listen on its own IP-address. Make sure to add its IP-address in the beginning on the line. (running as a cluster currently requires this due to a bug, so we will do this although we are not setting up a cluster).

bind 192.168.1.62 127.0.0.1

Enable redis to listen on the network, instead of only listening on the loopback device.

protected-mode no

For better dataintegrity enable this.

appendonly yes

Now restart the redis master

service redis-server restart

And stop the replicas

service redis-server stop

On each of the replicas, set them up af a replica of the master. (This was earlier called slaveof)

replicaof 192.168.1.62 6379

Start the replicas again

services redis-server start

You should now be able to see the replicas from the master redis-cli -h 192.168.1.62.

192.168.1.62:6379> info replication
# Replication
role:master
connected_slaves:2
slave0:ip=192.168.1.63,port=6379,state=online,offset=2615,lag=0
slave1:ip=192.168.1.64,port=6379,state=online,offset=2615,lag=1
master_replid:79a21f08669303e71990ea8819830f30c94384c6
master_replid2:0000000000000000000000000000000000000000
master_repl_offset:2615
second_repl_offset:-1
repl_backlog_active:1
repl_backlog_size:1048576
repl_backlog_first_byte_offset:1
repl_backlog_histlen:2615

Redis Sentinel

Stop the sentinel

service redis-sentinel stop

Change the contents of the redis configuration file in /etc/redis/sentinel.conf.

Add the IP-address for each of the servers. Each of the redis servers needs to listen on its own IP-address.

bind 192.168.1.62 127.0.0.1

Configure each of the hosts to monitor the redis master server, and at least 2 sentinels should agree to change the master.

sentinel monitor mymaster 192.168.1.62 6379 2

Set the time the redis master can be down

sentinel down-after-milliseconds mymaster 5000

Set the number of replicas who can change master at the same time. Setting this too high may cause none of the replicas to respond.

sentinel parallel-syncs mymaster 1

Finally set the timeout before the failover happens

sentinel failover-timeout mymaster 10000

Start the sentinel

service redis-sentinel start

-  ⇤ ← Revision 36 as of 2021-01-04 21:45:49 → 
  Size: 3538
  Editor: Kristian Kallenberg
  Comment:
+   ← Revision 56 as of 2021-03-01 22:47:08 → ⇥
  Size: 3990
  Editor: Kristian Kallenberg
  Comment:
-Deletions are marked like this.
+Additions are marked like this.
 Line 5:
- * 192.168.1.59 redis01
 * 192.168.1.60 redis02
 * 192.168.1.61 redis03
+ * 192.168.1.61 redis (virtual IP-address)
 * 192.168.1.62 redis01 (master)
 * 192.168.1.63 redis02 (replica)
 * 192.168.1.64 redis03 (replica)
-Line 21:
+Line 22:
-Add the IP-address for each of the servers. Each of the redis servers needs to listen on its own IP-address.
+Add the IP-address for each of the servers. Each of the redis servers needs to listen on its own IP-address. Make sure to add its IP-address in the beginning on the line. (running as a cluster currently requires this due to a bug, so we will do this although we are not setting up a cluster).
-Line 23:
+Line 24:
-bind 192.168.1.58 127.0.0.1 ::1
+bind 192.168.1.62 127.0.0.1
-Line 31:
+Line 32:
-Redis stores its data in memory only, hence its speed. To make sure redis does not loose its state, in case of a hardware failure, this will enable it to store the state as it changes in a file on disk.
+For better dataintegrity enable this.
-Line 36:
+Line 37:
-Activate redis cluster features
{{{
cluster-enabled yes
cluster-config-file nodes-6379.conf
cluster-node-timeout 15000
}}}

Restart redis on all hosts.
+Now restart the redis master
-Line 48:
+Line 42:
-Next we will need to configure the servers to run as a cluster
+And stop the replicas
-Line 50:
+Line 44:
-redis-cli --cluster create 192.168.1.59:6379 192.168.1.60:6379 192.168.1.61:6379
+service redis-server stop
-Line 53:
+Line 47:
-You should get output similar to this
+On each of the replicas, set them up af a replica of the master. (This was earlier called slaveof)
-Line 55:
+Line 49:
->>> Performing hash slots allocation on 3 nodes...
Master[0] -> Slots 0 - 5460
Master[1] -> Slots 5461 - 10922
Master[2] -> Slots 10923 - 16383
M: 78252519c58c5e7651554b3c82ba5800e4c11c41 192.168.1.59:6379
   slots:[0-5460] (5461 slots) master
M: b26d5a1ebe4891c68acf1d6b7cea5cc19057849f 192.168.1.60:6379
   slots:[5461-10922] (5462 slots) master
M: 0db96b401cca249b61660ed4d2945d658dcdc6e1 192.168.1.61:6379
   slots:[10923-16383] (5461 slots) master
Can I set the above configuration? (type 'yes' to accept): yes
>>> Nodes configuration updated
>>> Assign a different config epoch to each node
>>> Sending CLUSTER MEET messages to join the cluster
Waiting for the cluster to join

>>> Performing Cluster Check (using node 192.168.1.59:6379)
M: 78252519c58c5e7651554b3c82ba5800e4c11c41 192.168.1.59:6379
   slots:[0-5460] (5461 slots) master
M: 0db96b401cca249b61660ed4d2945d658dcdc6e1 192.168.1.61:6379
   slots:[10923-16383] (5461 slots) master
M: b26d5a1ebe4891c68acf1d6b7cea5cc19057849f 192.168.1.60:6379
   slots:[5461-10922] (5462 slots) master
[OK] All nodes agree about slots configuration.
>>> Check for open slots...
>>> Check slots coverage...
[OK] All 16384 slots covered.
+replicaof 192.168.1.62 6379
-Line 84:
+Line 52:
+Start the replicas again
{{{
services redis-server start
}}}

You should now be able to see the replicas from the master `redis-cli -h 192.168.1.62`.
{{{
192.168.1.62:6379> info replication
# Replication
role:master
connected_slaves:2
slave0:ip=192.168.1.63,port=6379,state=online,offset=2615,lag=0
slave1:ip=192.168.1.64,port=6379,state=online,offset=2615,lag=1
master_replid:79a21f08669303e71990ea8819830f30c94384c6
master_replid2:0000000000000000000000000000000000000000
master_repl_offset:2615
second_repl_offset:-1
repl_backlog_active:1
repl_backlog_size:1048576
repl_backlog_first_byte_offset:1
repl_backlog_histlen:2615
}}}

=== Redis Sentinel ===

Stop the sentinel
{{{
service redis-sentinel stop
}}}

Change the contents of the redis configuration file in `/etc/redis/sentinel.conf`.

Add the IP-address for each of the servers. Each of the redis servers needs to listen on its own IP-address.
{{{
bind 192.168.1.62 127.0.0.1
}}}

Configure each of the hosts to monitor the redis master server, and at least 2 sentinels should agree to change the master.
{{{
sentinel monitor mymaster 192.168.1.62 6379 2
}}}

Set the time the redis master can be down
{{{
sentinel down-after-milliseconds mymaster 5000
}}}

Set the number of replicas who can change master at the same time. Setting this too high may cause none of the replicas to respond.
{{{
sentinel parallel-syncs mymaster 1
}}}

Finally set the timeout before the failover happens
{{{
sentinel failover-timeout mymaster 10000
}}}

Start the sentinel
{{{
service redis-sentinel start
}}}
-Line 98:
+Line 127:
+ * https://www.willandskill.se/en/setup-a-highly-available-redis-cluster-with-sentinel-and-haproxy/

Redis

Software

Configuration

Redis Server

Redis Sentinel

References