ssdb cluster + keepalived build actual combat-3.ssdb master-standby mode build
surroundings
Operating system: CentOS Linux release 7.6.1810 (Core)
ssdb: 1.9.7
IP:
master: 10.80.2.121
slave: 10.80.2.85
vip：10.80.2.156
Kernel file modification
Be sure to remember to modify your Linux kernel parameters. For the content of max open files (maximum file descriptors), please refer to Building C1000K Server (1)-Basic
Use the c1000k tool to test how many concurrent connections your system supports: c1000k download address
1. Download and install
Download software:
[root@localhost ssdb]# cd /opt/ssdb/
[root@localhost ssdb]# wget --no-check-certificate https://github.com/ideawu/c1000k/archive/master.zip
Unzip the installation package:
[root@localhost ssdb]# unzip master.zip.1
Archive: master.zip.1
f435ee41fca847b71aab679fed1d9b07c979cfba
creating: c1000k-master/
extracting: c1000k-master/.gitignore
inflating: c1000k-master/Makefile
inflating: c1000k-master/README.md
inflating: c1000k-master/client.c
inflating: c1000k-master/server.c
Resolution:
[root@localhost ssdb]# cd c1000k-master
[root@localhost c1000k-master]# make
gcc -std=c99 -O2 -o server server.c
gcc -O2 -o client client.c
2. Check the maximum number of server connections
Start the service, at ports 1000 to 1099:
[root@localhost c1000k-master]# ./server 1000
Run the client on another client:
[root@localhost ~]# cd /opt/ssdb/c1000k-master/
[root@localhost c1000k-master]# ls
client client.c Makefile README.md server server.c
[root@localhost c1000k-master]# ./client 127.0.0.1 1000
connections: 922
error: Connection refused
At this time, the server displays:
connections: 921
error: Too many open files
The server can accept a maximum of 921 connections
ssdb basic configuration
Refer to the official documentation: ssdb configuration
Modify the /opt/data/ssdb.conf file:
1. Monitor
The default configuration is:
server:
ip: 127.0.0.1
port: 8888
This will only accept local connections and change to:
server:
#ip: 127.0.0.1
port: 8888
# bind to public ip
ip: 0.0.0.0
# format: allow|deny: all|ip_prefix
# multiple allows or denys is supported
#deny: all
allow: 127.0.0.1
allow: 10.80
allow: 10.82
Such admixing 10.80.*.*and 10.82.*.*connecting segment, such as the network is not limited, and only open 0.0.0.0to
2. Log
ssdb-server will output such a log every 5 minutes. The
default configuration is:
logger:
level: debug
output: log.txt
rotate:
size: 1000000000
- logger.level The log level. Supported log levels are: debug, info, warn, error, fatal. In general, it is recommended that you set logger.level to debug level.
- logger.output Log output. You can write the relative path or absolute path directly. If the relative path is the directory where the relative configuration file is located.
3.LevelDB
-
leveldb.cache_size Memory cache size, unit MB. Generally, the larger the number, the better the performance. You can set it to half of the physical memory. If your machine has less memory, then change it to a smaller value. The minimum value is 16.
-
leveldb.block_size don't care
-
leveldb.write_buffer_size write buffer size in MB. If your machine's memory is small, then change it to small, otherwise change it. It should be within this range: [4, 128];
-
leveldb.compaction_speed, under normal circumstances, don't care. If your hard drive performance is very poor, at the same time, your data hardly changes, and there is no new data written, you can change it to smaller (preferably greater than 50).
-
leveldb.compression compresses the data on the hard disk, it is best to set it to yes! If yes, generally you can store data of 10 times the hard disk space, and the performance will be better.
Memory calculation:
The memory occupied by an ssdb-server instance is instantaneous (it is possible, and even if it is reached, it only lasts for a short time). The maximum is (MB):
cache_size + write_buffer_size * 66 + 32
This is the case when the compression option is not turned on. If compression: yes, the calculation formula is:
cache_size + 10 * write_buffer_size * 66 + 32
You can adjust the configuration parameters to limit the memory usage of ssdb-server.
For an instance with a general load, the continuous occupation of physical memory is:
cache_size + write_buffer_size * 4 + 32
According to actual experience, using the default configuration instance will occupy about 1GB of memory. You can refer to this experience.
4. My main library configuration:
Memory resources:
total used free shared buff/cache available
Mem: 3.9G 149M 3.5G 24M 210M 3.5G
Swap: 2.0G 0B 2.0G
ssbd configuration:
[root@ecloud02-carchat-prod-ssdb01 data]# cat /opt/data/ssdb.conf
# ssdb-server config
# MUST indent by TAB!
# absolute path, or relative to path of this file, directory must exists
work_dir = ./var
pidfile = ./var/ssdb.pid
server:
#ip: 127.0.0.1
port: 8888
# bind to public ip
ip: 0.0.0.0
# format: allow|deny: all|ip_prefix
# multiple allows or denys is supported
#deny: all
allow: 127.0.0.1
allow: 10.80
allow: 10.82
# auth password must be at least 32 characters
#auth: very-strong-password
#readonly: yes
# in ms, to log slowlog with WARN level
#slowlog_timeout: 5
replication:
binlog: yes
# Limit sync speed to *MB/s, -1: no limit
sync_speed: -1
slaveof:
# to identify a master even if it moved(ip, port changed)
# if set to empty or not defined, ip:port will be used.
#id: svc_2
# sync|mirror, default is sync
#type: sync
#host: localhost
#port: 8889
logger:
level: debug
output: log.txt
rotate:
size: 1000000000
leveldb:
# in MB
cache_size: 2000
# in MB
write_buffer_size: 64
# in MB/s
compaction_speed: 1000
# yes|no
compression: yes
Note: The configuration file of SSDB uses a TAB to indicate first-level indentation, do not use spaces to indent, whether you use 1, 2, 3, 4, 5, 6, 7, 8, or Countless spaces will not work!
ssdb master-slave configuration
Refer to the official documentation: ssdb synchronization and replication configuration and monitoring
1. Modify the database configuration
What I need to configure here is: Master-slave mode, only need to modify the configuration of the slave library side:
replication:
binlog: yes
# Limit sync speed to *MB/s, -1: no limit
sync_speed: -1
slaveof:
# to identify a master even if it moved(ip, port changed)
# if set to empty or not defined, ip:port will be used.
id: svc_1
# sync|mirror, default is sync
type: sync
host: 10.80.2.121
port: 8888
Specified for the synchronous copy replication (Sync), and the host-based configuration repository ip
For the read-only, which can be modified from the library server.readonly: yesto
Then restart the slave ssdb service to make the configuration take effect:
[root@ecloud02-carchat-prod-ssdb02 data]# systemctl restart ssdb
2. Master-slave state
Enter the main library and check the status:
[root@ecloud02-carchat-prod-ssdb01 data]# /opt/data/ssdb-cli -h 10.80.2.121 -p 8888
ssdb (cli) - ssdb command line tool.
Copyright (c) 2012-2016 ssdb.io
'h' or 'help' for help, 'q' to quit.
ssdb-server 1.9.7
ssdb 10.80.2.121:8888> info
version
1.9.7
links
1
total_calls
6
dbsize
258
binlogs
capacity : 20000000
min_seq : 1
max_seq : 4
replication
client 10.80.2.85:13093
type : sync
status : SYNC
last_seq : 4
serv_key_range
kv : "" - ""
hash: "" - ""
zset: "" - ""
list: "" - ""
data_key_range
kv : "" - ""
hash: "" - ""
zset: "" - ""
list: "" - ""
leveldb.stats
Compactions
Level Files Size(MB) Time(sec) Read(MB) Write(MB)
--------------------------------------------------
0 1 0 0 0 0
19 result(s) (0.001 sec)
(0.001 sec)
From the library state:
[root@ecloud02-carchat-prod-ssdb01 data]# /opt/data/ssdb-cli -h 10.80.2.85 -p 8888
ssdb (cli) - ssdb command line tool.
Copyright (c) 2012-2016 ssdb.io
'h' or 'help' for help, 'q' to quit.
ssdb-server 1.9.7
ssdb 10.80.2.85:8888> info
version
1.9.7
links
1
total_calls
1
dbsize
258
binlogs
capacity : 20000000
min_seq : 1
max_seq : 4
replication
slaveof 10.80.2.121:8888
id : svc_1
type : sync
status : SYNC
last_seq : 4
copy_count : 0
sync_count : 0
serv_key_range
kv : "" - ""
hash: "" - ""
zset: "" - ""
list: "" - ""
data_key_range
kv : "" - ""
hash: "" - ""
zset: "" - ""
list: "" - ""
leveldb.stats
Compactions
Level Files Size(MB) Time(sec) Read(MB) Write(MB)
--------------------------------------------------
0 1 0 0 0 0
19 result(s) (0.001 sec)
(0.001 sec)
- links
The current number of server connections.
- total_calls
The number of requests processed since the server was started.
- dbsize
The estimated size of the database (may be very different from the hard disk occupancy, as shown by the du command), the number of bytes. If the server has compression enabled, this size is the compressed size.
- binlogs
The write status of the current instance.
capacity: binlog 队列的最大长度
min_seq: 当前队列中的最小 binlog 序号
max_seq: 当前队列中的最大 binlog 序号
- replication
There can be multiple replication records. Each one represents a connected slave (client), or a master (slaveof) connected to the current server.
slaveof|client host:port, 远端 master/slave 的 host:port.
type: 类型, sync|mirror.
status: 当前同步状态, DISCONNECTED|INIT|OUT_OF_SYNC|COPY|SYNC.
last_seq: 上一条发送或者收到的 binlog 的序号.
slaveof.id: master 的 id(这是从 slave's 角度来看的, 你永远不需要在 master 上配置它自己的 id).
slaveof.copy_count: 在全量同步时, 已经复制的 key 的数量.
slaveof.sync_count: 发送或者收到的 binlog 的数量.
- Where status:
DISCONNECTED: 与 master 断开了连接, 一般是网络中断.
INIT: 初始化状态.
OUT_OF_SYNC: 由于短时间内在 master 有大量写操作, 导致 binlog 队列淘汰, slave 丢失同步点, 只好重新复制全部的数据.
COPY: 正在复制基准数据的过程中, 新的写操作可能无法及时地同步.
SYNC: 同步状态是健康的.
- key_range.*
The keys of different data types are sorted in SSDB, so this information indicates the minimum key and maximum key of different data types.
- leveldb.stats
This information shows the number of files and the total file size of each level of LevelDB. The smaller the level, the fewer the files, the healthier the database (the faster the query).
3. Determine the synchronization status
For the master, binlogs.max_seq refers to the sequence number of the latest write (write / update / delete) operation on the current instance, and replication.client.last_seq refers to the sequence number of the latest binlog that has been sent to the slave.
Therefore, if you want to determine whether the master-slave synchronization has been synchronized in place (real-time update), then determine whether binlogs.max_seq and replication.client.last_seq are equal, and the status is SYNC.
Experiment: The
main library writes data:
[root@ecloud02-carchat-prod-ssdb01 data]# /opt/data/ssdb-cli -h 10.80.2.121 -p 8888
ssdb (cli) - ssdb command line tool.
Copyright (c) 2012-2016 ssdb.io
'h' or 'help' for help, 'q' to quit.
ssdb-server 1.9.7
ssdb 10.80.2.121:8888> set a 123
ok
(0.001 sec)
ssdb 10.80.2.121:8888> q
bye.
View data from the library:
[root@ecloud02-carchat-prod-ssdb01 data]# /opt/data/ssdb-cli -h 10.80.2.85 -p 8888
ssdb (cli) - ssdb command line tool.
Copyright (c) 2012-2016 ssdb.io
'h' or 'help' for help, 'q' to quit.
ssdb-server 1.9.7
ssdb 10.80.2.85:8888> get a
123
(0.001 sec)
Successful verification!
PS
If you use flushdb to delete all data, the master and slave will be invalid. In the latest version of ssdb, flushdb has been disabled in master-slave mode:
ssdb 10.80.2.121:8888> flushdb
error! - flushdb is not allowed when replication is in use!
