Grokbase Groups HBase user June 2016
FAQ
Our regionservers are periodically seg faulting, roughly once a day, and I
would appreciate some help debugging. Some version details:

Java HotSpot(TM) 64-Bit Server VM (24.80-b11) for linux-amd64 JRE
(1.7.0_80-b15), built on Apr 10 2015 19:53:14 by "java_re" with gcc 4.3.0
20080428 (Red Hat 4.3.0-8)

hbase-regionserver 0.98.6+cdh5.3.2+83-1.cdh5.3.2.p0.17~precise-cdh5.3.2

The segfault log output is:
https://gist.github.com/hazzadous/aa5013f50824658e75b75fd860c73f02

This has details of the java options as well. Heap is ~46GB using G1GC.

The log output from HBase around this event is:
https://gist.github.com/hazzadous/4b94b5dcb351a360881c04fb54b5a70f

I'm going to attempt to update to java 8 to see if there's any improvements
but any analysis will be appreciated.

Harry

Search Discussions

  • Esteban Gutierrez at Jun 10, 2016 at 5:08 pm
    Hi Harry,

    As you mentioned, moving to JDK8 is a good idea. There are many known
    issues with G1GC and JDK7 that make using the G1 collector unreliable and
    you will see that kind of crashes once in a while.

    cheers,
    esteban.


    --
    Cloudera, Inc.

    On Fri, Jun 10, 2016 at 9:35 AM, Harry Waye wrote:

    Our regionservers are periodically seg faulting, roughly once a day, and I
    would appreciate some help debugging. Some version details:

    Java HotSpot(TM) 64-Bit Server VM (24.80-b11) for linux-amd64 JRE
    (1.7.0_80-b15), built on Apr 10 2015 19:53:14 by "java_re" with gcc 4.3.0
    20080428 (Red Hat 4.3.0-8)

    hbase-regionserver 0.98.6+cdh5.3.2+83-1.cdh5.3.2.p0.17~precise-cdh5.3.2

    The segfault log output is:
    https://gist.github.com/hazzadous/aa5013f50824658e75b75fd860c73f02

    This has details of the java options as well. Heap is ~46GB using G1GC.

    The log output from HBase around this event is:
    https://gist.github.com/hazzadous/4b94b5dcb351a360881c04fb54b5a70f

    I'm going to attempt to update to java 8 to see if there's any improvements
    but any analysis will be appreciated.

    Harry
  • Harry Waye at Jun 15, 2016 at 6:46 pm
    This seems to have resolve the issue, no sigsegv seen... yet.

    Thanks Esteban

    Harry
    On Fri, Jun 10, 2016 at 6:08 PM Esteban Gutierrez wrote:

    Hi Harry,

    As you mentioned, moving to JDK8 is a good idea. There are many known
    issues with G1GC and JDK7 that make using the G1 collector unreliable and
    you will see that kind of crashes once in a while.

    cheers,
    esteban.


    --
    Cloudera, Inc.

    On Fri, Jun 10, 2016 at 9:35 AM, Harry Waye wrote:

    Our regionservers are periodically seg faulting, roughly once a day, and I
    would appreciate some help debugging. Some version details:

    Java HotSpot(TM) 64-Bit Server VM (24.80-b11) for linux-amd64 JRE
    (1.7.0_80-b15), built on Apr 10 2015 19:53:14 by "java_re" with gcc 4.3.0
    20080428 (Red Hat 4.3.0-8)

    hbase-regionserver 0.98.6+cdh5.3.2+83-1.cdh5.3.2.p0.17~precise-cdh5.3.2

    The segfault log output is:
    https://gist.github.com/hazzadous/aa5013f50824658e75b75fd860c73f02

    This has details of the java options as well. Heap is ~46GB using G1GC.

    The log output from HBase around this event is:
    https://gist.github.com/hazzadous/4b94b5dcb351a360881c04fb54b5a70f

    I'm going to attempt to update to java 8 to see if there's any
    improvements
    but any analysis will be appreciated.

    Harry

Related Discussions

Discussion Navigation
viewthread | post
Discussion Overview
groupuser @
categorieshbase, hadoop
postedJun 10, '16 at 4:36p
activeJun 15, '16 at 6:46p
posts3
users2
websitehbase.apache.org

2 users in discussion

Harry Waye: 2 posts Esteban Gutierrez: 1 post

People

Translate

site design / logo © 2018 Grokbase