Redis is often referred as a data structures server. What this means is that Redis provides access to mutable data structures via a set of commands, which are sent using a server-client model with TCP sockets and a simple protocol. So different processes can query and modify the same data structures in a shared way.
Data structures implemented into Redis have a few special properties:
Redis cares to store them on disk, even if they are always served and modified into the server memory. This means that Redis is fast, but that is also non-volatile.
Implementation of data structures stress on memory efficiency, so data structures inside Redis will likely use less memory compared to the same data structure modeled using an high level programming language.
Redis offers a number of features that are natural to find in a database, like replication, tunable levels of durability, cluster, high availability.
Another good example is to think of Redis as a more complex version of memcached, where the operations are not just SETs and GETs, but operations to work with complex data types like Lists, Sets, ordered data structures, and so forth.
If you want to know more, this is a list of selected starting points:
################################## INCLUDES ################################### # 这在你有标准配置模板但是每个 redis 服务器又需要个性设置的时候很有用。 include /path/to/local.conf include /path/to/other.conf
################################ GENERAL #####################################
# 设置 1 或另一个设置为 0 禁用这个特性。 # Setting one or the other to 0 disables the feature. # By default min-slaves-to-write is set to 0 (feature disabled) and # min-slaves-max-lag is set to 10.
To configure a master-slave redis with optimum performance, here are some suggestions:
Kernel Configure:
Make sure to set the Linux kernel overcommit memory setting to 1. Add vm.overcommit_memory = 1 to /etc/sysctl.conf and then reboot or run the command sysctl vm.overcommit_memory=1 for this to take effect immediately.This make redis to consider server always have enough memory to fork.
Make sure to disable Linux kernel feature transparent huge pages, it will affect greatly both memory usage and latency in a negative way. This is accomplished with the following command: echo never > /sys/kernel/mm/transparent_hugepage/enabled. It will change memory page size from 4kb to 2M, increase memory usage on copy-on-write. https://docs.mongodb.com/manual/tutorial/transparent-huge-pages/#red-hat-centos-7
Common redis:
maxclients 50000
tcp-keepalive 300, enabled to check tcp dead connections
daemonize yes, not using systemd to manage
maxmemory=70% of physical memory, leave enough safe space for forking to do background save or full replication. Since remain memory can be used for accumulate changes(which cause copy-on-write) during background save or AOF rewrite.
stop-writes-on-bgsave-error no, to allow client write even on failure of rdb background save, increase robustness of redis master.
repl-backlog-size 1gb. should be determined by the write LOAD of clients, the default is 1mb, we can raise it to 1gb.
repl-backlog-ttl 0,never release backlog buffer
client-output-buffer-limit slave 0 0, disable soft limit, set hardlimit same as maxmemory size, if set this to unlimited, and slave is blocking, will used up all memory and force master to evict all keys.
maxmemory-policy volatile-lru, only volatile keys will be evicted.
appendfsync everysec，if both master and slave crash due to power shortage, we can ensure only lost 1 second data in slave.
no-appendfsync-on-rewrite yes, make sure slave would be not blocking by rewrite process if try to write AOF log, so slave can follow master closely, we prefer availability to durability.
aof-load-truncated yes, let redis fix truncated error by itself.
The repl_backlog is only allocated once there is at least a replica connected.
If slave disconnected from master and write changes exceed repl-backlog-size. The master will do a background save, which will block all clients when forking which utilize copy-on-write to allocated memory if changed during dump, it may not double the memory usage if memory page size is small and load not too high.
No need to enable repl-diskless-sync , it also requires fork, but only will write to socket instead of disk.
can change python logic code to use hash type for kv type, such as redis tokens, it can reduce memory profoundly
hash-max-ziplist-value 64
hash-max-ziplist-entries 512
Redis master
save “”
appendonly no
disable RDB and AOF, since process forking caused by RDB and AOF will block all commands. INFO command to check fork time for each gigabytes: latest_fork_usec:2568287us / used_memory_human: 76.02G =33.784ms, for each gigabyte, it may take 33ms to fork
Redis slave:
enable RDB dump.
save 900 1 save 300 10 save 60 10000
enable AOF：
appendonly yes
be cautious if master is empty, the slaveof can wipe out all slave keys.
Fail-over
config set save “” -- disable RDB in master
config set appendonly no -- disable AOF in master
take note on latest_fork_usec:5934556, this means redis master will be blocked for 5.9 seconds, and if keepalived healthy check mistakenly consider this as downtime, its VIP will failover to slave which cause the current syncing slave become master, in this case, do not empty slave when trying to sync with master which may cause empty ‘master‘ at the moment.
When slave try to sync with master, remember to set keepalived fail count to be much bigger number, to prevent keepalived falsely to do unwanted failover.
If AOF enabled, redis will ignore dump.rdb file during initial loading.
On instance, where only one redis, can use service keepalived restart to failover, but if with two or more instance, it has to stop_redis.sh to let keepalived failover to another instance
Never stop_redis.sh on master, it may lost all data if appendonly.aof is empty, and cause long init loading time if data set is bigger!
[root@sg-gop-10-71-12-78 redis-6389]# cat check_redis.sh #!/bin/bash # Check if redis is running, return 1 if not. # Used by keepalived to initiate a failover in case redis is down
REDIS_STATUS=$(telnet 127.0.0.1 6389 < /dev/null | grep "Connected" ) if [ "$REDIS_STATUS" != "" ] then exit 0 else logger "REDIS is NOT running. Setting keepalived state to FAULT." exit 1 fi
