这几天在搭Cassandra集群以及对Cassandra的性能测试,步骤还挺多,记录一下。
关于Caaandra在服务器上配置多节点集群,可以参考一下文章:
http://blog.csdn.net/cloud_xy/article/details/48091003
http://blog.csdn.net/cloud_xy/article/details/48107251
以及Cassandra官方文档:http://cassandra.apache.org/doc/latest/getting_started/configuring.html
注意:最好关掉集群中每台服务器的防火墙,确保相应端口通过,以便节点之间能互相访问。
systemctl stop firewalld.service #停止firewall
systemctl disable firewalld.service #禁止firewall开机启动
firewall-cmd --state #查看默认防火墙状态(关闭后显示notrunning,开启后显示running)
Cassandra的具体使用请参见Cassandra日常运维:http://zqhxuyuan.github.io/2015/10/15/Cassandra-Daily/
和官方Documentation:http://cassandra.apache.org/doc/latest/getting_started/index.html
关于使用YCSB压测Cassandra 3.11.2:
准备工作·:最好仔细阅读YCSB的wiki:https://github.com/brianfrankcooper/YCSB/wiki
环境:三台服务器:ip 200.200.172.117-119
Cassandra3.11.2
1.在https://github.com/brianfrankcooper/YCSB/获取源代码,解压至本地目录即可。
2.用cassandra的cqlsh创建keyspace和cloumn family
(1)新建keyspace:
cqlsh> create keyspace usertable with replication = {'class':'SimpleStrategy', 'replication_factor':3};
注意:数据复制有两种策略:
-
SimpleStrategy
:仅用于单个数据中心和一个机架。如果您打算使用多个数据中心,请使用NetworkTopologyStrategy
。 -
NetworkTopologyStrategy
:强烈建议用于大多数部署,因为未来扩展需要扩展到多个数据中心时更容易。
replication_factor:复制因子。如果class是SimpleStrategy,则是必需的; 否则,不使用。多个节点上数据的复制数量。
复制因子1意味着一个节点上每行只有一个副本。复制因子2意味着每行的两个副本,其中每个副本位于不同的节点上。所有复制品都同样重要; 没有主要或主要副本。作为一般规则,复制因子不应超过群集中的节点数量。
(2)应用keyspace:
cqlsh> USE usertable;
(3)新建table也即cloumn family:
create table usertable (y_id varchar primary key,field0 varchar,field1 varchar,field2 varchar,field3 varchar,field4 varchar,field5 varchar,field6 varchar,field7 varchar,field8 varchar,field9 varchar);
3、查看ycsb命令格式:
#cd bin
#ycsb
usage: ./ycsb command database [options] Commands:
load Execute the load phase
run Execute the transaction phase
shell Interactive mode Databases:
accumulo https://github.com/brianfrankcooper/YCSB/tree/master/accumulo
aerospike https://github.com/brianfrankcooper/YCSB/tree/master/aerospike
arangodb https://github.com/brianfrankcooper/YCSB/tree/master/arangodb
asynchbase https://github.com/brianfrankcooper/YCSB/tree/master/asynchbase
basic https://github.com/brianfrankcooper/YCSB/tree/master/basic
cassandra-cql https://github.com/brianfrankcooper/YCSB/tree/master/cassandra
cassandra2-cql https://github.com/brianfrankcooper/YCSB/tree/master/cassandra2
couchbase https://github.com/brianfrankcooper/YCSB/tree/master/couchbase
couchbase2 https://github.com/brianfrankcooper/YCSB/tree/master/couchbase2
dynamodb https://github.com/brianfrankcooper/YCSB/tree/master/dynamodb
elasticsearch https://github.com/brianfrankcooper/YCSB/tree/master/elasticsearch
geode https://github.com/brianfrankcooper/YCSB/tree/master/geode
googlebigtable https://github.com/brianfrankcooper/YCSB/tree/master/googlebigtable
googledatastore https://github.com/brianfrankcooper/YCSB/tree/master/googledatastore
hbase094 https://github.com/brianfrankcooper/YCSB/tree/master/hbase094
hbase098 https://github.com/brianfrankcooper/YCSB/tree/master/hbase098
hbase10 https://github.com/brianfrankcooper/YCSB/tree/master/hbase10
hypertable https://github.com/brianfrankcooper/YCSB/tree/master/hypertable
infinispan https://github.com/brianfrankcooper/YCSB/tree/master/infinispan
infinispan-cs https://github.com/brianfrankcooper/YCSB/tree/master/infinispan
jdbc https://github.com/brianfrankcooper/YCSB/tree/master/jdbc
kudu https://github.com/brianfrankcooper/YCSB/tree/master/kudu
mapkeeper https://github.com/brianfrankcooper/YCSB/tree/master/mapkeeper
memcached https://github.com/brianfrankcooper/YCSB/tree/master/memcached
mongodb https://github.com/brianfrankcooper/YCSB/tree/master/mongodb
mongodb-async https://github.com/brianfrankcooper/YCSB/tree/master/mongodb
nosqldb https://github.com/brianfrankcooper/YCSB/tree/master/nosqldb
orientdb https://github.com/brianfrankcooper/YCSB/tree/master/orientdb
rados https://github.com/brianfrankcooper/YCSB/tree/master/rados
redis https://github.com/brianfrankcooper/YCSB/tree/master/redis
riak https://github.com/brianfrankcooper/YCSB/tree/master/riak
s3 https://github.com/brianfrankcooper/YCSB/tree/master/s3
solr https://github.com/brianfrankcooper/YCSB/tree/master/solr
tarantool https://github.com/brianfrankcooper/YCSB/tree/master/tarantool
voldemort https://github.com/brianfrankcooper/YCSB/tree/master/voldemort Options:
-P file Specify workload file
-cp path Additional Java classpath entries
-jvm-args args Additional arguments to the JVM
-p key=value Override workload property
-s Print status to stderr
-target n Target ops/sec (default: unthrottled)
-threads n Number of client threads (default: ) Workload Files:
There are various predefined workloads under workloads/ directory.
See https://github.com/brianfrankcooper/YCSB/wiki/Core-Properties
for the list of workload properties.
ycsb: error: too few arguments
从命令格式里可以看出 -P可以加载一些配置文件 -p可以以键值对的方式加载一些配置 -s每隔一段时间输出执行信息 -threads线程数
4、新建cassandra连接文件(里面的属性可以在源码https://github.com/brianfrankcooper/YCSB/blob/master/cassandra/src/main/java/com/yahoo/ycsb/db/CassandraCQLClient.java中查看)
#vim cassandra.properties hosts = spark131,spark130,spark129 #host列表,用逗号,隔开
port =
cassandra.keyspace = usertable #测试表
cassandra.username = ershixiong #cassandra用户名
cassandra.password = #cassandra密码
cassandra.readconsistencylevel = ANY
cassandra.writeconsistencylevel = ANY
cassandra.maxconnections =
cassandra.connecttimeoutmillis =
cassandra.readtimeoutmillis =
cassandra.properties这个文件没有的话就新建一个
注意:关于读写一致性级别的设置请参见文档:https://docs.datastax.com/en/archived/cassandra/2.0/cassandra/dml/dml_config_consistency_c.html
5、配置workload
#vim workloads/workloada
workload=com.yahoo.ycsb.workloads.CoreWorkload
readallfields=false
readproportion=0.5
updateproportion=0.5
scanproportion=
insertproportion=
requestdistribution=zipfian
fieldcount 表示每条数据中的字段数,默认为 ;
fieldlength 表示每个字段的值的长度,默认为 ;
readallfields 域用来标识是否读取所有的所有的字段,取值有 ture 或 false;
readproportion,
updateproportion,
scanproportion,
insertproportion 分别表示该 workload中读、更新、扫描和插入操作占总操作的百分比,这四个值的和为 ;
requestdistribution 表示数据的分布情况,当前支持 uniform,zipfian 和 latest,默认为 uniform;
maxscanlength 域主要为扫描操作定义,定义了最大扫描的记录数量,默认为 ;
scanlengthdistribution 域也是为扫描操作定义的,为每次扫描的长度定义相应的分布,默认是 uniform;
insertorder 域主要分两种 ordered 和 hashed,默认为 hashed;
operationcount 总共的 operation 数量;
maxexecutiontime 为该 workload 定义了最长的执行时间,单位为 s。
AverageLatency(平均潜伏期)平均潜伏期(average latency):指当磁头移动到数据所在的磁道后,然后等待所要的数据块继续转动(半圈或多些、少些)到磁头下的时间,单位为毫秒(ms)。平均潜伏期是越小越好,潜伏期小代表硬盘的读取数据的等待时间短,这就等于具有更高的硬盘数据传输率。
注意:还有一个变量:zeropadding可以设置Key的长度,具体workloads参数请参考YCSB源码:https://github.com/brianfrankcooper/YCSB/blob/master/core/src/main/java/com/yahoo/ycsb/workloads/CoreWorkload.java
关于workload的不同类型说明(a、b、c、d)请参见官方文档:https://github.com/brianfrankcooper/YCSB/wiki/Core-Workloads
5、测试
YCSB测试包括两个阶段,Load the data 和 Run the workload
load阶段:
bin/ycsb load cassandra2-cql -P workloads/workloada -P cassandra.properties -p columnfamily=usertable -s -threads > load_32threads.dat
run阶段:
bin/ycsb run cassandra2-cql -P workloads/workloada -P cassandra.properties -p columnfamily=usertable -s -threads > run_32threads.dat
数据接口名为cassandra2-cql
加载ycsb的配置:workloads/workloada
加载cassandra的配置:cassandra.properties
columnfamily名称为usertable
32个线程执行
具体输出信息意义请参见官方文档:https://github.com/brianfrankcooper/YCSB/wiki/Running-a-Workload
常见错误:
1.All host(s) tried for query failed (tried: /ip (com.datastax.driver.core.TransportException:[/ip] Cannot connect), /ip (com.datastax.driver.core.TransportException: [/ip] Cannot connect))
解决:如果需要远程连接,那么cassandra.yaml里面的rpc_address就需要改变为您当前环境实际的IP地址,否则会报错!修改后就可以了!
2.SLF4J: Failed to load class "org.slf4j.impl.StaticLoggerBinder".
SLF4J: Defaulting to no-operation (NOP) logger implementation
SLF4J: See http://www.slf4j.org/codes.html#StaticLoggerBinder for further details.
解决:下载最新的slf 并将slf4j-simple-1.7.7.jar和slf4j-api-1.7.7.jar复制到ycsb的lib目录:https://www.slf4j.org/download.html