hbase GC造成数据处理延迟(GC pool)

心已赠人 2021-11-02 00:02 187阅读 0赞

时间压测过程中，通过程序加时间戳判断，在批量查询hbase的时候，会有部分数据查询时间大于8s，这肯定是不能接受的，查看hbase的日志发现原因,有类似日志'GC pool',推测是GC问题，各种查询资料：

> ...  
> 2018-03-01 17:32:16,243 WARN org.apache.hadoop.hbase.util.JvmPauseMonitor: Detected pause in JVM or host machine (eg GC): pause of approximately 16100ms  
> GC pool 'ParNew' had collection(s): count=1 time=0ms  
> GC pool 'ConcurrentMarkSweep' had collection(s): count=1 time=8791ms
> 
> ...........

解决：

参考<<Hbase不睡觉>>这本书JVM GC调优解决，在配置文件下面设置GC策略，增大内存，解决了此问题。

>  
> 
> \# Configure PermSize. Only needed in JDK7. You can safely remove it for JDK8+  
> \#export HBASE\_MASTER\_OPTS="$HBASE\_MASTER\_OPTS -XX:PermSize=128m -XX:MaxPermSize=512m"  
> \#export HBASE\_REGIONSERVER\_OPTS="$HBASE\_REGIONSERVER\_OPTS -XX:PermSize=256m -XX:MaxPermSize=256m"  
> \#export HBASE\_MASTER\_OPTS="$HBASE\_MASTER\_OPTS -Xmx512m"  
> \#export HBASE\_REGIONSERVER\_OPTS="$HBASE\_REGIONSERVER\_OPTS -Xmx512m"
> 
>  
> 
> \# export HBASE\_JMX\_BASE="-Dcom.sun.management.jmxremote.ssl=false -Dcom.sun.management.jmxremote.authenticate=false"  
> \# export HBASE\_MASTER\_OPTS="$HBASE\_MASTER\_OPTS $HBASE\_JMX\_BASE -Dcom.sun.management.jmxremote.port=10101"  
> \# export HBASE\_REGIONSERVER\_OPTS="$HBASE\_REGIONSERVER\_OPTS $HBASE\_JMX\_BASE -Dcom.sun.management.jmxremote.port=10102"  
> \# export HBASE\_THRIFT\_OPTS="$HBASE\_THRIFT\_OPTS $HBASE\_JMX\_BASE -Dcom.sun.management.jmxremote.port=10103"  
> \# export HBASE\_ZOOKEEPER\_OPTS="$HBASE\_ZOOKEEPER\_OPTS $HBASE\_JMX\_BASE -Dcom.sun.management.jmxremote.port=10104"  
> \# export HBASE\_REST\_OPTS="$HBASE\_REST\_OPTS $HBASE\_JMX\_BASE -Dcom.sun.management.jmxremote.port=10105"
> 
> \# File naming hosts on which HRegionServers will run.  $HBASE\_HOME/conf/regionservers by default.  
> \# export HBASE\_REGIONSERVERS=$\{HBASE\_HOME\}/conf/regionservers
> 
> \# Uncomment and adjust to keep all the Region Server pages mapped to be memory resident  
> \#HBASE\_REGIONSERVER\_MLOCK=true  
> \#HBASE\_REGIONSERVER\_UID="hbase"
> 
> \# File naming hosts on which backup HMaster will run.  $HBASE\_HOME/conf/backup-masters by default.  
> \# export HBASE\_BACKUP\_MASTERS=$\{HBASE\_HOME\}/conf/backup-masters
> 
> \# Extra ssh options.  Empty by default.  
> \# export HBASE\_SSH\_OPTS="-o ConnectTimeout=1 -o SendEnv=HBASE\_CONF\_DIR"
> 
> \# Where log files are stored.  $HBASE\_HOME/logs by default.  
> \# export HBASE\_LOG\_DIR=$\{HBASE\_HOME\}/logs
> 
> \# Enable remote JDWP debugging of major HBase processes. Meant for Core Developers   
> \# export HBASE\_MASTER\_OPTS="$HBASE\_MASTER\_OPTS -Xdebug -Xrunjdwp:transport=dt\_socket,server=y,suspend=n,address=8070"  
> \# export HBASE\_REGIONSERVER\_OPTS="$HBASE\_REGIONSERVER\_OPTS -Xdebug -Xrunjdwp:transport=dt\_socket,server=y,suspend=n,address=8071"  
> \# export HBASE\_THRIFT\_OPTS="$HBASE\_THRIFT\_OPTS -Xdebug -Xrunjdwp:transport=dt\_socket,server=y,suspend=n,address=8072"  
> \# export HBASE\_ZOOKEEPER\_OPTS="$HBASE\_ZOOKEEPER\_OPTS -Xdebug -Xrunjdwp:transport=dt\_socket,server=y,suspend=n,address=8073"
> 
> \# A string representing this instance of hbase. $USER by default.  
> \# export HBASE\_IDENT\_STRING=$USER
> 
> \# The scheduling priority for daemon processes.  See 'man nice'.  
> \# export HBASE\_NICENESS=10
> 
> \# The directory where pid files are stored. /tmp by default.  
> \# export HBASE\_PID\_DIR=/var/hadoop/pids
> 
> \# Seconds to sleep between slave commands.  Unset by default.  This  
> \# can be useful in large clusters, where, e.g., slave rsyncs can  
> \# otherwise arrive faster than the master can service them.  
> \# export HBASE\_SLAVE\_SLEEP=0.1
> 
> \# Tell HBase whether it should manage it's own instance of Zookeeper or not.  
> \# export HBASE\_MANAGES\_ZK=true
> 
> \# The default log rolling policy is RFA, where the log file is rolled as per the size defined for the   
> \# RFA appender. Please refer to the log4j.properties file to see more details on this appender.  
> \# In case one needs to do log rolling on a date change, one should set the environment property  
> \# HBASE\_ROOT\_LOGGER to "<DESIRED\_LOG LEVEL>,DRFA".  
> \# For example:  
> \# HBASE\_ROOT\_LOGGER=INFO,DRFA  
> \# The reason for changing default to RFA is to avoid the boundary case of filling out disk space as   
> \# DRFA doesn't put any cap on the log size. Please refer to HBase-5655 for more context.
> 
> export JAVA\_HOME=/java/jdk  
> export HBASE\_MANAGES\_ZK=false  
>

也可以参考：[https://blog.csdn.net/xiefu5hh/article/details/53090725][https_blog.csdn.net_xiefu5hh_article_details_53090725]

[https_blog.csdn.net_xiefu5hh_article_details_53090725]: https://blog.csdn.net/xiefu5hh/article/details/53090725