Predixy 中文版
Predixy is a high performance and fully featured proxy for redis sentinel and redis cluster
- High performance and lightweight.
- Multi-threads support.
- Works on Linux, OSX, BSD, Windows(Cygwin).
- Supports Redis Sentinel, single/multi redis group[s].
- Supports Redis Cluster.
- Supports redis block command, eg:blpop, brpop, brpoplpush.
- Supports scan command, even multi redis instances.
- Multi-keys command support: mset/msetnx/mget/del/unlink/touch/exists.
- Multi-databases support, means redis command select is avaliable.
- Supports redis transaction, limit in Redis Sentinel single redis group.
- Supports redis Scripts, script load, eval, evalsha.
- Supports redis Pub/Sub.
- Multi-DataCenters support, read from slaves.
- Extend AUTH, readonly/readwrite/admin permission, keyspace limit.
- Log level sample, async log record.
- Log file auto rotate by time and/or file size.
- Stats info, CPU/Memory/Requests/Responses and so on.
- Latency monitor.
Predixy can be compiled and used on Linux, OSX, BSD, Windows(Cygwin). Requires C++11 compiler.
It is as simple as:
$ make
To build in debug mode:
$ make debug
Some other build options:
- CXX=c++compiler, default is g++, you can specify other, eg:CXX=clang++
- EV=epoll|poll|kqueue, default it is auto detect according by platform.
- MT=false, disable multi-threads support.
- TS=true, enable predixy function call time stats, debug only for developer.
For examples:
$ make CXX=clang++
$ make EV=poll
$ make MT=false
$ make debug MT=false TS=true
Just copy src/predixy to the install path
$ cp src/predixy /path/to/bin
See below files:
- predixy.conf, basic config, will refrence below config files.
- cluster.conf, Redis Cluster backend config.
- sentinel.conf, Redis Sentinel backend config.
- auth.conf, authority control config.
- dc.conf, multi-datacenters config.
- latency.conf, latency monitor config.
$ src/predixy conf/predixy.conf
With default predixy.conf, Predixy will listen at 0.0.0.0:7617 and proxy to Redis Cluster 127.0.0.1:6379. In general, 127.0.0.1:6379 is not running in Redis Cluster mode. So you will look mass log output, but you can still test it with redis-cli.
$ redis-cli -p 7617 info
More command line arguments:
$ src/predixy -h
Like redis, predixy use INFO command to give stats.
Show predixy running info and latency monitors
redis> INFO
Show latency monitors by latency name
redis> INFO Latency <latency-name>
A latency monitor example:
LatencyMonitorName:all
latency(us) sum(us) counts
<= 100 3769836 91339 91.34%
<= 200 777185 5900 97.24%
<= 300 287565 1181 98.42%
<= 400 185891 537 98.96%
<= 500 132773 299 99.26%
<= 600 85050 156 99.41%
<= 700 85455 133 99.54%
<= 800 40088 54 99.60%
<= 1000 67788 77 99.68%
> 1000 601012 325 100.00%
T 60 6032643 100001
The last line is total summary, 60 is average latency(us)
Show latency monitors by server address and latency name
redis> INFO ServerLatency <server-address> [latency-name]
Reset all stats and latency monitors, require admin permission.
redis> CONFIG ResetStat
- Local Redis instance running on default port (6379)
- C++11 compiler
- Make
- Use Homebrew to install Redis in a single instance mode with cluster mode disabled to test your changes locally
$ brew install redis
- Start Redis
$ brew services start redis
- Verify your local Redis is running:
$ redis-cli -h 127.0.0.1 -p 6379 ping
Should return "PONG"
$ redis-cli -h 127.0.0.1 -p 6379 info
Will return info about the Redis instance, if required for debugging
Note, logs for Redis installed by Brew will appear in /opt/homebrew/var/log/redis.log
by default.
- Compile Predixy:
$ make
This will create object files and the executable in the src
directory.
- Verify your local Redis is running:
$ redis-cli -h 127.0.0.1 -p 6379 ping
Should return "PONG"
- Start Predixy using the local configuration:
$ src/predixy conf/predixy_local.conf
You should see output indicating Predixy is listening on 127.0.0.1:7617
- Test the connection through Predixy:
# Basic connectivity test
$ redis-cli -h 127.0.0.1 -p 7617 ping
# Check Predixy status
$ redis-cli -h 127.0.0.1 -p 7617 info
# Test read/write operations
$ redis-cli -h 127.0.0.1 -p 7617 set test "Hello via Predixy"
$ redis-cli -h 127.0.0.1 -p 7617 get test
- To stop Predixy:
$ pkill -f predixy
- Create directories for each Redis node (we'll use 3 nodes):
$ mkdir -p redis-cluster/{7000,7001,7002}
- Create Redis configuration for each node. First for port 7000:
$ cat > redis-cluster/7000/redis.conf << EOL
port 7000
cluster-enabled yes
cluster-config-file nodes-7000.conf
cluster-node-timeout 5000
appendonly yes
dir ./
bind 127.0.0.1
daemonize no
EOL
- Copy and adjust configuration for other nodes:
$ cp redis-cluster/7000/redis.conf redis-cluster/7001/redis.conf
$ cp redis-cluster/7000/redis.conf redis-cluster/7002/redis.conf
$ sed -i '' 's/7000/7001/g' redis-cluster/7001/redis.conf
$ sed -i '' 's/7000/7002/g' redis-cluster/7002/redis.conf
- Start each Redis instance (in separate terminal windows):
$ redis-server redis-cluster/7000/redis.conf
$ redis-server redis-cluster/7001/redis.conf
$ redis-server redis-cluster/7002/redis.conf
- Create the cluster:
$ redis-cli --cluster create 127.0.0.1:7000 127.0.0.1:7001 127.0.0.1:7002 --cluster-replicas 0
Type 'yes' when prompted to accept the configuration.
- Verify cluster status:
$ redis-cli -p 7000 cluster nodes
- Start Predixy with the cluster configuration:
$ src/predixy conf/predixy_cluster.conf
- Test the cluster setup through Predixy:
# Basic connectivity test
$ redis-cli -p 7617 set test "Hello Cluster"
$ redis-cli -p 7617 get test
# Check Predixy cluster status
$ redis-cli -p 7617 info
When done testing, you can:
- Stop Predixy (Ctrl+C or
pkill predixy
) - Stop each Redis instance (Ctrl+C in each terminal)
- Remove cluster files:
rm -rf redis-cluster
Note: The cluster setup distributes keys across all nodes using hash slots. Each node in this setup handles approximately 5461 hash slots (16384/3 slots per node).
- Predixy will be listening on port 7617 while your Redis instance remains on 6379
- The configuration in
predixy_local.conf
is set up for a single local Redis instance - Build artifacts (*.o files) are ignored by git but can be safely kept for development
- Logs will appear in stdout by default
predixy is fast, how fast? more than twemproxy, codis, redis-cerberus
See wiki benchmark
Copyright (C) 2017 Joyield, Inc. <joyield.com#gmail.com>
All rights reserved.
License under BSD 3-clause "New" or "Revised" License