[ https://issues.apache.org/jira/browse/HBASE-1648?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13508579#comment-13508579 ]
liang xie commented on HBASE-1648: ---------------------------------- [~saint....@gmail.com] yes, it should be safe. we should not use G1 before JDK7u4 And also, i tried in a hbase0.94.0 + jdk6u22 + G1GC test env(not a good choice, aha): hbase(main):003:0> status 'simple' 2 live servers ha3:60020 1354522267193 requestsPerSecond=0, numberOfOnlineRegions=1, usedHeapMB=38, maxHeapMB=200 ha2:60020 1354551060862 requestsPerSecond=0, numberOfOnlineRegions=6, usedHeapMB=88, maxHeapMB=200 0 dead servers Aggregate load: 0, regions: 7 the output is fine And, i tried with jdk7u9 + G1GC as well: hbase(main):002:0> status 'simple' 2 live servers ha2:60020 1354552532489 requestsPerSecond=0, numberOfOnlineRegions=5, usedHeapMB=46, maxHeapMB=200 ha3:60020 1354523738930 requestsPerSecond=0, numberOfOnlineRegions=2, usedHeapMB=44, maxHeapMB=200 0 dead servers Aggregate load: 0, regions: 7 still good > heap usage and limit reporting doesn't work as expected when using G1 GC > ------------------------------------------------------------------------ > > Key: HBASE-1648 > URL: https://issues.apache.org/jira/browse/HBASE-1648 > Project: HBase > Issue Type: Bug > Environment: Linux Centos x86_64 5.3, Sun JDK 1.6.0_14 (HotSpot > 64-Bit Server VM build 14.0-b16) > Reporter: Andrew Purtell > Priority: Minor > > Getting some early experience with the G1 GC. > Running with -Xmx1000m and HBASE_OPTS set to "-XX:+UnlockExperimentalVMOpts > -XX:+UseG1GC". Regionserver heap use reports are not of much use: > {noformat} > hbase(main):001:0> status 'simple' > 3 live servers > test3:60020 1247389283042 > requests=0, regions=1, usedHeap=0, maxHeap=41 > test2:60020 1247389219994 > requests=0, regions=1, usedHeap=0, maxHeap=41 > test4:60020 1247389324563 > requests=0, regions=2, usedHeap=0, maxHeap=41 > 0 dead servers > {noformat} > top is about right: > {noformat} > PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND > 23988 hadoop 24 0 1492m 98m 10m S 0.0 2.5 0:02.65 java > {noformat} > Incidentally, don't try -XX:+UseG1GC and -XX:+DoEscapeAnalysis together or > the JVM will rapidly segfault. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira