redis-cluster-tool is a convenient and useful tool for redis cluster. It was primarily built to manage the redis cluster.
redis-cluster-tool need hiredis-vip(https://github.com/vipshop/hiredis-vip), so you must install hiredis-vip first.
To build redis-cluster-tool:
$ export HIREDIS_VIP_INSTALL_DIR="your hiredis-vip install path"
$ export PREFIX="the path for redis-cluster-tool to install"
$ make
$ sudo make install
You can "yum install redis-cluster-tool" or "apt-get install redis-cluster-tool" instead of building from source code.
Before install the package, execute the follow command first:
deb package : curl -s https://packagecloud.io/install/repositories/deep/packages/script.deb.sh | sudo bash
rpm package : curl -s https://packagecloud.io/install/repositories/deep/packages/script.rpm.sh | sudo bash
You can also download the packages from "https://packagecloud.io/deep/packages" and install by yourself.
Attention : package redis-cluster-tool is depend on package hiredis-vip.
If you want to support other OS packages, please contact with me.
Usage: redis-cluster-tool [-?hVds] [-v verbosity level] [-o output file]
[-c conf file] [-a addr] [-i interval]
[-p pid file] [-C command] [-r redis role]
[-t thread number] [-b buffer size]
Options:
-h, --help : this help
-V, --version : show version and exit
-d, --daemonize : run as a daemon
-s, --simple : show the output not in detail
-v, --verbosity=N : set logging level (default: 5, min: 0, max: 11)
-o, --output=S : set logging file (default: stderr)
-c, --conf-file=S : set configuration file (default: conf/rct.yml)
-a, --addr=S : set redis cluster address (default: 127.0.0.1:6379)
-i, --interval=N : set interval in msec (default: 1000 msec)
-p, --pid-file=S : set pid file (default: off)
-C, --command=S : set command to execute (default: cluster_state)
-r, --role=S : set the role of the nodes that command to execute on (default: all, you can input: all, master or slave)
-t, --thread=N : set how many threads to run the job(default: 8)
-b, --buffer=S : set buffer size to run the job (default: 1048576 byte, unit:G/M/K)
Commands:
cluster_state :Show the cluster state.
cluster_used_memory :Show the cluster used memory.
cluster_keys_num :Show the cluster holds keys num.
slots_state :Show the slots state.
node_slot_num :Show the node hold slots number.
new_nodes_name :Show the new nodes name that not covered slots.
cluster_rebalance :Show the cluster how to rebalance.
flushall :Flush all the cluster.
cluster_config_get :Get config from every node in the cluster and check consistency.
cluster_config_set :Set config to every node in the cluster.
cluster_config_rewrite :Rewrite every node config to echo node for the cluster.
node_list :List the nodes
del_keys :Delete keys in the cluster. The keys must match a given glob-style pattern.(This command not block the redis)
Now -d, -v, -o, -c, -i and -p options are not used, maybe they would be using in the future.
The command must be covered by double quotation marks, if there were more than one arguments.
Get the cluster state:
$redis-cluster-tool -a 127.0.0.1:34501 -C cluster_state -r master
master[127.0.0.1:34504] cluster_state is ok
master[127.0.0.1:34501] cluster_state is ok
master[127.0.0.1:34502] cluster_state is ok
master[127.0.0.1:34503] cluster_state is ok
all nodes cluster_state is ok
Get the cluster used memory:
$redis-cluster-tool -a 127.0.0.1:34501 -C cluster_used_memory -r master
master[127.0.0.1:34504] used 195 M 25%
master[127.0.0.1:34501] used 195 M 25%
master[127.0.0.1:34502] used 195 M 25%
master[127.0.0.1:34503] used 195 M 25%
cluster used 780 M
Rebalance the cluster slots:
$redis-cluster-tool -a 127.0.0.1:34501 -C cluster_rebalance
--from e1a4ba9922555bfc961f987213e3d4e6659c9316 --to 785862477453bc6b91765ffba0b5bc803052d70a --slots 2048
--from 437c719f50dc9d0745032f3b280ce7ecc40792ac --to cb8299390ce53cefb2352db34976dd768708bf64 --slots 2048
--from a497fc619d4f6c93bd4afb85f3f8a148a3f35adb --to a0cf6c1f12d295cd80f5811afab713cdc858ea30 --slots 2048
--from 0bdef694d08cb3daab9aac518d3ad6f8035ec896 --to 471eaf98ff43ba9a0aadd9579f5af1609239c5b7 --slots 2048
Then you can use "redis-trib.rb reshard --yes --from e1a4ba9922555bfc961f987213e3d4e6659c9316 --to 785862477453bc6b91765ffba0b5bc803052d70a --slots 2048 127.0.0.1:34501" to rebalance the cluster slots
Flushall the cluster:
$redis-cluster-tool -a 127.0.0.1:34501 -C flushall -s
Do you really want to execute the "flushall"?
please input "yes" or "no" :
yes
OK
Get a config from every node in cluster:
$redis-cluster-tool -a 127.0.0.1:34501 -C "cluster_config_get maxmemory" -r master
master[127.0.0.1:34501] config maxmemory is 1048576000 (1000MB)
master[127.0.0.1:34502] config maxmemory is 1048576000 (1000MB)
master[127.0.0.1:34503] config maxmemory is 1048576000 (1000MB)
master[127.0.0.1:34504] config maxmemory is 1048576000 (1000MB)
All nodes config are Consistent
cluster total maxmemory: 4194304000 (4000MB)
Set a config from every node in cluster:
$redis-cluster-tool -a 127.0.0.1:34501 -C "cluster_config_set maxmemory 10000000" -s
Do you really want to execute the "cluster_config_set"?
please input "yes" or "no" :
yes
OK
Delete keys in the cluster:
$redis-cluster-tool -a 127.0.0.1:34501 -C "del_keys 1*"
Do you really want to execute the "del_keys"?
please input "yes" or "no" :
yes
delete keys job is running...
delete keys job finished, deleted: 999999 keys, used: 4 s
You can also use '-t' to set the total threads for delete keys job.
Here we show a performance test:
-
45 redis nodes in cluster
-
24 threads to run the delete job(cpu has 24 threads)
-
Nine hundred million keys in the redis cluster
-
delete two hundred million keys
del_keys job used 92 seconds(about two million keys per second)
The del_keys job can be speed up with lock-free list in the future.
Copyright 2012 Deep, Inc.
Licensed under the Apache License, Version 2.0: http://www.apache.org/licenses/LICENSE-2.0