Cassandra2.0 JNA配置

Cassandra中通过JNA可以调用Native方法优化Jvm内存，大概意思就是不让操作系统将jvm内存置换到系统swap分区，从而提升jvm内存性能。它有两个用处：

 

• Since 0.6.2: JNA for mlockall. This prevents Linux from swapping out parts of the JVM that aren't accessed frequently. Chris Goffinet reported a 13% performance improvement in his tests from this change. CASSANDRA-1214

• Since 0.6.6: JNA for hard links, improving snapshots. Previously Cassandra would use the /bin/ln binary to create hard links to SSTables when creating a snapshot. On nodes with thousands of SSTables, this would take a very long time, because it had to fork+exec the JVM to run the new process. With JNA, Cassandra uses the link call directly. CASSANDRA-1371

P.S. JNA配置不是必须的，但官方建议生成环境下使用JNA。

 

我的软件环境：CentOS6 64bit + Cassandra2.0(DataStax社区版，通过YUM安装的)

 

开启方法如下：

 

1. 下载并放置JNA

wget https://maven.java.net/content/repositories/releases/net/java/dev/jna/jna/4.0.0/jna-4.0.0.jar
cp jna-4.0.0.jar /usr/share/cassandra/lib/

 

2. 开启JNA

 

vi /etc/security/limits.d/cassandra.conf

cassandra soft memlock unlimited
cassandra hard memlock unlimited

注意： 通过yum安装的dsc是通过cassandra用户启动的，所以我配置上面为cassandra用户。

 

3. 验证安装

如果没有配置JNA，cassandra log将会打印如下信息：

 

INFO 12:17:58,616 JNA not found. Native methods will be disabled.

 

 配置成功后，会出现如下log：

INFO 09:59:59,184 JNA mlockall successful

 

4. 错误诊断

将JNA jar包放置后，但如果没有配置好系统limit的话，将会出现如下log：

WARN 10:49:53,587 Unable to lock JVM memory (ENOMEM). This can result in part of the JVM being swapped out, especially with mmapped I/O enabled. Increase RLIMIT_MEMLOCK or run Cassandra as root.

 完成步骤2配置，就可以解决这个警告。

 

参考：

https://journal.paul.querna.org/articles/2010/11/11/enabling-jna-in-cassandra/

 

--end