Preparation

• Create a DCS Proxy Cluster instance.

  The available memory of the instance must be larger than the size of the data to be imported.

• Prepare a server which will be used for transmitting and importing data.

  The server must be connected to the DCS instance and have sufficient disk space for storing migration data.

  You can use HUAWEI CLOUD ECS and configure the same VPC, subnet, and security group for the ECS and the DCS instance.

Procedure

1. Export the RDB file.

   Download and install Redis: http://download.redis.io/releases/redis-5.0.8.tar.gz.

   Run the following command to export the RDB file for each shard in the cluster:

   redis-cli -h {source_redis_address} -p {source_redis_port} -a {source_redis_password} --rdb {output.rdb}

   If "Transfer finished with success." is displayed after the command is executed, the file is exported successfully.

   

   1. Migrate data during off-peak hours.
   2. When exporting Codis data, individually export the data of each shard in the cluster, and then import the data file by file.

2. Upload the RDB file to the server. HUAWEI CLOUD ECS is used as an example.

   If the files are larger, compress them before transmission.

3. Download redis-port.

   Log in to the ECS. Run the following commands to download and decompress the tool package. redis-port can be used directly without compilation.

   wget https://github.com/CodisLabs/redis-port/releases/download/v2.0-beta/redis-port-v2.0-beta-go1.10.1-linux.tar.gz

   tar -xvf redis-port-v2.0-beta-go1.10.1-linux.tar.gz

4. Import data.

   Run the following commands in redis-port to import all RDB files to the target DCS instance. If the ECS and the DCS instance are in the same VPC, it takes 4 to 10 seconds to export 1 million data records, assuming that each record contains 20 byte data.

   redis-restore -n {N} -i {outputfile.rdb} -t {password}@{dcs_instance_address}:{port} [--unixtime-in-milliseconds="yyyy-MM-dd hh:mm:ss"]

   By specifying the db parameter, you can import the cached data of the specified DB in the file. -n indicates that multiple CPUs process the import task concurrently.

   Parameter description:

   root@redis-nodelete:~/port/redis-port-v2.0-beta-go1.10.1-linux# ./redis-restore  --help
   Usage:
           redis-restore [--ncpu=N] [--input=INPUT|INPUT] --target=TARGET [--aof=FILE] [--db=DB] [--unixtime-in-milliseconds=EXPR]
           redis-restore  --version
    
   Options:
           -n N, --ncpu=N                    Set runtime.GOMAXPROCS to N.
           -i INPUT, --input=INPUT           Set input rdb encoded file.
           -t TARGET, --target=TARGET        The target redis instance ([auth@]host:port).
           -a FILE, --aof=FILE               Also restore the replication backlog.
           --db=DB                           Accept db = DB, default is *.
           --unixtime-in-milliseconds=EXPR   Update expire time when restoring objects from RDB.
    
   Examples:
           $ redis-restore    dump.rdb -t 127.0.0.1:6379
           $ redis-restore -i dump.rdb -t 127.0.0.1:6379 --aof dump.aof --db=1
           $ redis-restore             -t 127.0.0.1:6379 --aof dump.aof
           $ redis-restore             -t 127.0.0.1:6379 --db=0
           $ redis-restore -i dump.rdb -t 127.0.0.1:6379 --unixtime-in-milliseconds="@209059200000"       // ttlms += (now - '1976-08-17')
           $ redis-restore -i dump.rdb -t 127.0.0.1:6379 --unixtime-in-milliseconds="+1000"               // ttlms += 1s
           $ redis-restore -i dump.rdb -t 127.0.0.1:6379 --unixtime-in-milliseconds="-1000"               // ttlms -= 1s
           $ redis-restore -i dump.rdb -t 127.0.0.1:6379 --unixtime-in-milliseconds="1976-08-17 00:00:00" // ttlms += (now - '1976-08-17')

   Example:

   root@redis-nodelete:~/port/redis-port-v2.0-beta-go1.10.1-linux# ./redis-restore -i save196.rdb -t ******@192.168.0.171:6379
   2018/03/26 09:15:33 restore.go:70: [INFO] restore: input = "save196.rdb", aoflog = "" target = "******@192.168.0.171:6379"
   2018/03/26 09:15:33 restore.go:126: [INFO] restore: (r,f,s/a,f,s) = (rdb,rdb.forward,rdb.skip/aof,rdb.forward,rdb.skip)
   2018/03/26 09:15:34 restore.go:155: [INFO] restore: size = 46721058 - [ 49.94%,  0.00%]   (r,f,s/a,f,s)=(23330816,0,599496/0,0,0)     ~  (22.25mb,-,-/0,-,-)
   2018/03/26 09:15:35 restore.go:155: [INFO] restore: size = 46721058 - [ 99.31%,  0.00%]   (r,f,s/a,f,s)=(46399488,12558,1179884/0,0,0)    ~  (44.25mb,-,-/0,-,-)
   2018/03/26 09:15:35 restore.go:155: [INFO] restore: size = 46721058 - [100.00%,  0.00%]   (r,f,s/a,f,s)=(46721058,20000,1179884/0,0,0)    ~  (44.56mb,-,-/0,-,-)
   2018/03/26 09:15:35 restore.go:159: [INFO] restore: done
   root@redis-nodelete:~/port/redis-port-v2.0-alpha-go1.9.2-linux# 

5. Verify the migration.

   After the migration is completed, access the DCS instance and run the info command to check whether the data has been successfully imported as required.

   If the data import fails, run the flushall or flushdb command to clear the cached data in the instance, and import the data again.