FAQ
Hi HBaser,

we meet a problem when a meta region down then another region server load
region. a region server core.

out machine enviroment:
8 cores CPU, 48G memory, 12 1T disks

our region server heap size : 36000

jdk
java version "1.6.0_07"
Java(TM) SE Runtime Environment (build 1.6.0_07-b06)
Java HotSpot(TM) 64-Bit Server VM (build 10.0-b23, mixed mode)

our jvm parameter like below:
-XX:+UseConcMarkSweepGC -XX:+UseParNewGC -XX:+UseCMSCompactAtFullCollection
-XX:CMSFullGCsBeforeCompaction=0 -XX:+CMSClassUnloadingEnabled
-XX:-CMSParallelRemarkEnabled -XX:CMSInitiatingOccupancyFraction=70
-XX:SoftRefLRUPolicyMSPerMB=0 -XX:MaxTenuringThreshold=7
And I attach hs_err.log
Has anyone met this issue ??
or If I can get your jdk version and cluster environment is more
pleasure.


Thanks & Best regards
LiuJinglong

Search Discussions

  • Stack at Sep 26, 2010 at 4:43 am

    On Sat, Sep 25, 2010 at 9:32 PM, baggio liu wrote:
    Hi HBaser,
    we meet a problem when a meta region down then another region server load
    region. a region server core.
    It core dumps?


    out machine enviroment:
    8 cores CPU, 48G memory, 12 1T disks
    our region server heap size : 36000
    jdk
    java version "1.6.0_07"
    This is an old JVM. A lot of bugs have been fixed since this version
    was released. Can you update? Especially given you have so many
    cores and a bunch of RAM.



    Java(TM) SE Runtime Environment (build 1.6.0_07-b06)
    Java HotSpot(TM) 64-Bit Server VM (build 10.0-b23, mixed mode)
    our jvm parameter like below:
    -XX:+UseConcMarkSweepGC -XX:+UseParNewGC -XX:+UseCMSCompactAtFullCollection
    -XX:CMSFullGCsBeforeCompaction=0 -XX:+CMSClassUnloadingEnabled
    -XX:-CMSParallelRemarkEnabled -XX:CMSInitiatingOccupancyFraction=70
    -XX:SoftRefLRUPolicyMSPerMB=0 -XX:MaxTenuringThreshold=7
    This looks like an interesting configuration. How did you arrive at
    it (especially max tenuring threshold = 7 and
    cmsfulgcsbeforecompaction).


    And I attach hs_err.log
    Has anyone met this issue ??
    or If I can get your jdk version and cluster environment is more
    pleasure.
    At SU we run u14. u18 is generally regarded as bad. u20 may have
    issues. u21, the latest, purportedly has some improved heuristics to
    help w/ heap fragmentation.

    Try updating the jvm first.

    St.Ack

    Thanks & Best regards
    LiuJinglong



  • Baggio liu at Sep 26, 2010 at 5:17 am
    Thanks for your reply. Please see my comment inline.

    Thanks & Best regards
    LiuJinglong

    2010/9/26 Stack <stack@duboce.net>
    On Sat, Sep 25, 2010 at 9:32 PM, baggio liu wrote:
    Hi HBaser,
    we meet a problem when a meta region down then another region server load
    region. a region server core.
    It core dumps?
    Yes, it's core dump

    out machine enviroment:
    8 cores CPU, 48G memory, 12 1T disks
    our region server heap size : 36000
    jdk
    java version "1.6.0_07"
    This is an old JVM. A lot of bugs have been fixed since this version
    was released. Can you update? Especially given you have so many
    cores and a bunch of RAM.

    Sure, our HBase cluster has not released, we'll try to use another
    version jvm. (u14)
    Java(TM) SE Runtime Environment (build 1.6.0_07-b06)
    Java HotSpot(TM) 64-Bit Server VM (build 10.0-b23, mixed mode)
    our jvm parameter like below:
    -XX:+UseConcMarkSweepGC -XX:+UseParNewGC
    -XX:+UseCMSCompactAtFullCollection
    -XX:CMSFullGCsBeforeCompaction=0 -XX:+CMSClassUnloadingEnabled
    -XX:-CMSParallelRemarkEnabled -XX:CMSInitiatingOccupancyFraction=70
    -XX:SoftRefLRUPolicyMSPerMB=0 -XX:MaxTenuringThreshold=7
    This looks like an interesting configuration. How did you arrive at
    it (especially max tenuring threshold = 7 and
    cmsfulgcsbeforecompaction).
    In fact, we are not very clear effect of these configurations in
    performance . These config are tuning in our Hadoop cluster(namenode)
    according to experience.

    And I attach hs_err.log
    Has anyone met this issue ??
    or If I can get your jdk version and cluster environment is more
    pleasure.
    At SU we run u14. u18 is generally regarded as bad. u20 may have
    issues. u21, the latest, purportedly has some improved heuristics to
    help w/ heap fragmentation.

    Try updating the jvm first.
    Our production cluster (hadoop) use u7, and it seems stable. How do you
    choose u14, because some issues in jdk before ??
    What about u16 ?? I just see jdk community recommend this version.


    St.Ack

    Thanks & Best regards
    LiuJinglong



Related Discussions

Discussion Navigation
viewthread | post
Discussion Overview
groupuser @
categorieshbase, hadoop
postedSep 26, '10 at 4:33a
activeSep 26, '10 at 5:17a
posts3
users2
websitehbase.apache.org

2 users in discussion

Baggio liu: 2 posts Stack: 1 post

People

Translate

site design / logo © 2022 Grokbase