hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Joarder KAMAL <joard...@gmail.com>
Subject Re: Confusion with Actual Number of Regions in HBase
Date Thu, 21 Feb 2013 05:46:21 GMT
Thanks all. Actually the problem happened due to accidentally remove some
security policy which defines the ports to communicate between the servers
and required by ZK to detect region servers.
However, the problem leads me to the opportunity to know how to tune HBase
for long GC pauses as well as others.


Regards,
Joarder Kamal


On 21 February 2013 15:18, ramkrishna vasudevan <
ramkrishna.s.vasudevan@gmail.com> wrote:

> The two RS could have died due to GC or some connectivity issues with ZK.
>  Checking the logs would help us get a clear idea.
>
> REgards
> Ram
>
> On Thu, Feb 21, 2013 at 7:09 AM, Joarder KAMAL <joarderm@gmail.com> wrote:
>
> > Hi Ted,
> >
> > After login into the HBase UI I found two of my region servers were dead.
> > Don't have any clue why. Master is still up and running for the past 4
> > days.
> > Memory is still running pretty low without any workload and two region
> > servers!!
> >
> > --------------------------------------------------
> >
> > root@hdb1-01:/home/ubuntu# jps
> > 1064 NameNode
> > 1578 JobTracker
> > 10770 Jps
> > 2481 HRegionServer
> > 2261 HMaster
> > 1280 DataNode
> > 1769 TaskTracker
> > 2197 HQuorumPeer
> > 1480 SecondaryNameNode
> > --------------------------------------------------
> > root@hdb1-02:/home/ubuntu# jps
> > 1492 HRegionServer
> > 1315 TaskTracker
> > 1137 DataNode
> > 3290 Jps
> > --------------------------------------------------
> > root@hdb1-03:/home/ubuntu# jps
> > 1527 HRegionServer
> > 1171 DataNode
> > 7475 Jps
> > 1349 TaskTracker
> >
> > --------------------------------------------------
> > PID found 2197, 2261 and 2481 for 'hbase'
> > --------------------------------------------------
> > root@hdb1-01:/home/ubuntu# ps axo pid,ppid,rss,vsz | grep 2197
> >  2197     1 52248 1938732
> > --------------------------------------------------
> > root@hdb1-01:/home/ubuntu# ps axo pid,ppid,rss,vsz | grep 2261
> >  2261     1 108724 2007912
> > --------------------------------------------------
> > root@hdb1-01:/home/ubuntu# ps axo pid,ppid,rss,vsz | grep 2481
> >  2481     1 686252 2031376
> > --------------------------------------------------
> > root@hdb1-01:/home/ubuntu# free -m
> >              total       used       free     shared    buffers     cached
> > Mem:     1652       1616         36          0         51         305
> >
> > -/+ buffers/cache:       1259        393
> > Swap:          895          8        887
> >
> >
> > Thanks a lot again ...
> >
> >
> > Regards,
> > Joarder Kamal
> >
> >
> >
> > On 21 February 2013 11:08, Ted Yu <yuzhihong@gmail.com> wrote:
> >
> > > What do you see when you use 'ps aux | grep <pid>' command ?
> > > A sample is shown below:
> > >
> > > USER              PID  %CPU %MEM      VSZ    RSS   TT  STAT STARTED
> > >  TIME COMMAND
> > > tyu             16468 232.2  1.4  4974136 240528 s002  R+    4:05PM
> > > 2:11.15
> > >
> > >
> >
> /System/Library/Java/JavaVirtualMachines/1.6.0.jdk/Contents/Home/bin/javadoc
> > > -J-Xmx2048m @options @packages
> > >
> > > Please give us the VSZ and RSS readings.
> > >
> > > Thanks
> > >
> > > On Wed, Feb 20, 2013 at 3:56 PM, Joarder KAMAL <joarderm@gmail.com>
> > wrote:
> > >
> > > > Sorry may be I didn't clarify the question properly.
> > > >
> > > > Q2. As per the default setting maxHeapMB size is 991, so in
> saturation
> > > > condition HBase can't use more RAM (even if it is available, for
> > example
> > > > the system memory size is 4GB) than this defined size, is it right?
> > > >
> > > >
> > > > Many thanks again...
> > > >
> > > > On 21 February 2013 10:48, Ted Yu <yuzhihong@gmail.com> wrote:
> > > >
> > > > > I am not familiar with monitoring tools on EC2.
> > > > >
> > > > > You can use Ganglia to monitor hadoop / HBase.
> > > > >
> > > > > Cheers
> > > > >
> > > > > On Wed, Feb 20, 2013 at 3:42 PM, Joarder KAMAL <joarderm@gmail.com
> >
> > > > wrote:
> > > > >
> > > > > > Hi Ted,
> > > > > >
> > > > > > Thanks for the reply.
> > > > > >
> > > > > > Q1. Cleared now. Sorry I didn't count the -ROOT- and .META.
> tables
> > > > > > Q2. Not clear. Could explain a bit more? I am using the default
> > > > setting.
> > > > > >
> > > > > > Thanks again.
> > > > > >
> > > > > >
> > > > > > On 21 February 2013 10:36, Ted Yu <yuzhihong@gmail.com>
wrote:
> > > > > >
> > > > > > > bq. but when I click to view the table details it shows
there
> > are 5
> > > > > > > regions.
> > > > > > >
> > > > > > > Did you include -ROOT- and .META. tables above ?
> > > > > > >
> > > > > > > bq. as you can see the maxHeapMB size is 991
> > > > > > >
> > > > > > > Attachment is automatically stripped from your email. FYI
> > > > > > >
> > > > > > > On Wed, Feb 20, 2013 at 3:26 PM, Joarder KAMAL <
> > joarderm@gmail.com
> > > >
> > > > > > wrote:
> > > > > > >
> > > > > > > > Dear All,
> > > > > > > >
> > > > > > > > I am currently running a 3 node HBase cluster with
Amazon EC2
> > > small
> > > > > > > > instance (1CPU, 1.7GB Mem). After running a simple
workload
> > (type
> > > > A)
> > > > > > > using
> > > > > > > > YCSB benchmarking tool, I noticed in the Web UI that
the
> number
> > > of
> > > > > > > regions
> > > > > > > > in the 3 region servers are 7 but when I click to
view the
> > table
> > > > > > details
> > > > > > > it
> > > > > > > > shows there are 5 regions. I am a bit confuse. Does
the Web
> UI
> > > > > showing
> > > > > > > > wrong results? How can I check region details from
the
> shell/is
> > > > there
> > > > > > any
> > > > > > > > other command line tools for that? I am using Hbase
0.94.3
> (as
> > it
> > > > is
> > > > > > the
> > > > > > > > latest version compatible with YCSB at this moment)
with
> > > > > Hadoop-1.0.4.
> > > > > > > >
> > > > > > > > Another thing, as you can see the maxHeapMB size is
991
> > (default
> > > > > > setting)
> > > > > > > > then whenever I am running workloads the server free
memory
> > > status
> > > > is
> > > > > > > > terrible - almost 0MB for 1M/10M transactions
> > (YCSB/Workload-A).
> > > My
> > > > > > > > question is if the maxHeapMB is fixed at 1GB then
why I am
> > > loosing
> > > > > > extra
> > > > > > > > 700MB memory? How can I monitor cpu/memory/io usage
for
> > > individual
> > > > > > > cluster
> > > > > > > > components (HBase/Hadoop/MapRed/ZK, etc.)? Is there
any tool
> > > > > available
> > > > > > > with
> > > > > > > > HBase?
> > > > > > > >
> > > > > > > > Thanks a lot.
> > > > > > > >
> > > > > > > >
> > > > > > > > [image: Inline images 1]
> > > > > > > >
> > > > > > > >
> > > > > > > > Regards,
> > > > > > > > Joarder Kamal
> > > > > > > >
> > > > > > > > ------------------------------------------
> > > > > > > > PhD Student
> > > > > > > > Gippsland School of Information Technology,
> > > > > > > > Faculty of Information Technology,
> > > > > > > > MONASH University
> > > > > > > > Churchill VIC, Australia. 3842
> > > > > > > >
> > > > > > > >
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message