In many cases, the client application resides on the same machine as the server, so there is no need to listen do network sockets. If this is the case you may want to use unix sockets instead, for this you need to set the **port** option to 0, and then enable unix sockets with the following options.
Set the path to the socket file
unixsocket /tmp/redis.sock
Set restricted permission to the socket file
unixsocketperm 700
Now, to have access with redis-cli you should use the -s flag pointing to the socket file
redis-cli -s /tmp/redis.sock
#### requirepass ####
You may need remote access, if so, you should use a password, that will be required before any operation.
requirepass "bTFBx1NYYWRMTUEyNHhsCg"
#### rename-command ####
Imagine the output of the next command. Yes, it will dump the configuration of the server, so you should deny access to this kind information whenever is possible.
CONFIG GET *
To restrict, or even disable this and other commands by using the **rename-command**. You must provide a command name and a replacement. To disable, set the replacement string to "" (blank), this is more secure as it will prevent someone from guessing the command name.
Access through unix sockets with password and command changes
#### Snapshots ####
By default Redis will periodically dump its datasets to **dump.rdb** on the data directory we set. You can configure how often the rdb file will be updated by the save command, the first parameter is a timeframe in seconds and the second is a number of changes performed on the data file.
Every 15 hours if there was at least 1 key change
save 900 1
Every 5 hours if there was at least 10 key changes
save 300 10
Every minute if there was at least 10000 key changes
save 60 10000
The **/var/lib/redis/6379/dump.rdb** file contains a dump of the dataset on memory since last save. Since it creates a temporary file and then replace the original file, there is no problem of corruption and you can always copy it directly without fear.
### Starting at boot ###
You may use systemd to add Redis to the system startup
Copy sample init_script to /etc/init.d, note also the number of the port on the script name
cp utils/redis_init_script /etc/init.d/redis_6379
We are going to use systemd, so create a unit file named redis_6379.service under **/etc/systems/system**
vi /etc/systemd/system/redis_6379.service
Put this content, try man systemd.service for details
[Unit]
Description=Redis on port 6379
[Service]
Type=forking
ExecStart=/etc/init.d/redis_6379 start
ExecStop=/etc/init.d/redis_6379 stop
[Install]
WantedBy=multi-user.target
Now add the memory overcommit and maximum backlog options we have set before to the **/etc/sysctl.conf** file.
vm.overcommit_memory = 1
net.core.somaxconn=512
For the transparent huge pages support there is no sysctl directive, so you can put the command at the end of /etc/rc.local
echo never > /sys/kernel/mm/transparent_hugepage/enabled
### Conclusion ###
That's enough to start, with these settings you will be able to deploy Redis server for many simpler scenarios, however there is many options on redis.conf for more complex environments. On some cases, you may use [replication][4] and [Sentinel][5] to provide high availability, [split the data][6] across servers, create a cluster of servers. Thanks for reading!
