Hi Again,

One more symptom :  When I look the one of the  hbase table column ie:
PureText (which we configured to be LZO)  from the hadoop dfs , I can read
the file. Isnt it looks like a lzo'ed file ?  Is this shows LZO is not
worked ?  Or when I look the hbase LZO'ed file from dfs  its automatically
unlzo'ed to show me ?



On Sun, Dec 11, 2011 at 3:07 AM, lars hofhansl <[email protected]> wrote:

> You need enable LZO compression on the target table (the table you are
> importing to), but I assume you did that.
>
>
>
> ----- Original Message -----
> From: Lord Khan Han <[email protected]>
> To: [email protected]
> Cc:
> Sent: Saturday, December 10, 2011 10:09 AM
> Subject: Re: Hbase export / import Why doubling the Table Size ?
>
> When we exporting from hbase table which is LZO compression on it, the
> exported file is decompressed  or as is with LZO  columns?
>
>
>
> On Sat, Dec 10, 2011 at 6:40 PM, Lord Khan Han <[email protected]
> >wrote:
>
> > It is a succes for both lzo  snappy.  Content is the html document.. Web
> > document
> >
> >
> > hbase org.apache.hadoop.hbase.util.CompressionTest
> > hdfs://localhost:8020/user/root/testfile.lzo lzo
> >
> > 11/12/10 18:37:04 INFO lzo.GPLNativeCodeLoader: Loaded native gpl library
> >
> > 11/12/10 18:37:04 INFO lzo.LzoCodec: Successfully loaded & initialized
> > native-lzo library [hadoop-lzo rev
> 2ad6654f3e9cad97d13f716e51a0509253c0aabb]
> >
> > 11/12/10 18:37:04 INFO compress.CodecPool: Got brand-new compressor
> >
> > SUCCESS
> >
> >
> >
> >
> >
> > On Sat, Dec 10, 2011 at 1:03 PM, Lars George <[email protected]
> >wrote:
> >
> >> Could you use the ComressionTest to verify that the library path is set
> >> up properly?
> >>
> >> $ hbase org.apache.hadoop.hbase.util.CompressionTest
> >> hdfs://<your-namenode>:8020/<some-writable-path>/test.lzo lzo
> >>
> >> Does it report OK? Same for Snappy? The reason I am asking is that when
> >> it does not find the native libs it uses no compression at all, and if
> your
> >> original was compressed then you will see the copied one being
> uncompressed
> >> and therefore much larger.
> >>
> >> Also, what is the content like? How large are the cells that are stored?
> >>
> >> Lars
> >>
> >>
> >> On Dec 10, 2011, at 8:53 AM, Lord Khan Han wrote:
> >>
> >> > I will check the reverse export imprt to cdh3b4 today to see is it
> same
> >> > size in the cluster..
> >> >
> >> > when we use the hadoop dst copy how we candeal with the .META ?
> because
> >> we
> >> > are copying 1 tabel not all and also there is region info in .META
> >> > including their dns which is different offcoures in new  cluster.
> >> >
> >> > I tried the import again today with no compression.. It is doubled the
> >> > exported file size!!  I mean I have 200gig exported hbase table size.
> >> when
> >> > import without compression its going 400gig.. Its definitely writing
> >> twice
> >> > something..
> >> >
> >> > thanks
> >> >
> >> >
> >> >
> >> > On Sat, Dec 10, 2011 at 2:19 AM, lars hofhansl <[email protected]>
> >> wrote:
> >> >
> >> >> There's copytable (also an MR job - written by J-D), but it reuses
> the
> >> >> mapper class from the Import.java, so it
> >> >> probably won't make a difference.
> >> >>
> >> >> What I meant to say below... When you export/import the table from
> your
> >> >> CDH3u2 cluster back to your CDH3B4
> >> >> cluster, is the size still doubled?
> >> >>
> >> >>
> >> >> If both clusters are shutdown, you can use Hadoop's distcp to copy
> >> >> directly on the filesystem level; in fact that might be your
> >> >> best option.
> >> >>
> >> >> -- Lars
> >> >>
> >> >>
> >> >> ----- Original Message -----
> >> >> From: Lord Khan Han <[email protected]>
> >> >> To: [email protected]; lars hofhansl <[email protected]>
> >> >> Cc:
> >> >> Sent: Friday, December 9, 2011 4:05 PM
> >> >> Subject: Re: Hbase export / import Why doubling the Table Size ?
> >> >>
> >> >> Thanks for your time..
> >> >>
> >> >> Is there any reliable way to copy table between these cluster instead
> >> of
> >> >> export/import?
> >> >>
> >> >>
> >> >>
> >> >> On Sat, Dec 10, 2011 at 1:39 AM, lars hofhansl <[email protected]>
> >> >> wrote:
> >> >>
> >> >>> Hmm... I'm afraid I am out of options. If you want you can try to
> copy
> >> >> the
> >> >>> table
> >> >>> from CHD3u2 and your CDH3B4 system, and see if the size remains
> >> doubled.
> >> >>>
> >> >>> Does this happen with very small table, too? If so, you could take a
> >> >> small
> >> >>> sample
> >> >>> HFile and upload it (both the CHD3B4 and CDH3u2 versions) somewhere
> so
> >> >>> that we can have a look.
> >> >>>
> >> >>>
> >> >>> -- Lars
> >> >>>
> >> >>>
> >> >>> ----- Original Message -----
> >> >>> From: Lord Khan Han <[email protected]>
> >> >>> To: [email protected]; lars hofhansl <[email protected]>
> >> >>> Cc:
> >> >>> Sent: Friday, December 9, 2011 2:45 PM
> >> >>> Subject: Re: Hbase export / import Why doubling the Table Size ?
> >> >>>
> >> >>> in same configured cluster (carbon copy)  when I made import  there
> >> is no
> >> >>> increas on size.. same size..
> >> >>>
> >> >>> problem in the cdh3u2..
> >> >>>
> >> >>>
> >> >>> On Sat, Dec 10, 2011 at 12:42 AM, lars hofhansl <
> [email protected]>
> >> >>> wrote:
> >> >>>
> >> >>>> What happens when you export/import into the same (CDH3B4) cluster
> >> >> using
> >> >>> a
> >> >>>> new table name?
> >> >>>> Does the size double as well?
> >> >>>>
> >> >>>>
> >> >>>>
> >> >>>> ----- Original Message -----
> >> >>>> From: Lord Khan Han <[email protected]>
> >> >>>> To: [email protected]; lars hofhansl <[email protected]>
> >> >>>> Cc:
> >> >>>> Sent: Friday, December 9, 2011 2:27 PM
> >> >>>> Subject: Re: Hbase export / import Why doubling the Table Size ?
> >> >>>>
> >> >>>> I flush  ed  and major_compact  ed ..  nothing changed...   i am
> >> stuck
> >> >>> this
> >> >>>> last two days...:(  any idea?
> >> >>>>
> >> >>>>
> >> >>>> On Sat, Dec 10, 2011 at 12:11 AM, Lord Khan Han <
> >> >> [email protected]
> >> >>>>> wrote:
> >> >>>>
> >> >>>>> Now flushed  and compacting again..
> >> >>>>>
> >> >>>>> one more clue:
> >> >>>>>
> >> >>>>> I tested to import CDH3B4 (same as exported cluster) with lzo..
> all
> >> >> is
> >> >>>>> okay.. table size is same..
> >> >>>>> than I upgrade to cdh3u2  table also is ok and same size..
> >> >>>>>
> >> >>>>> But when I try to import in cdh3u2  this size doubling happens..
> >> >>>>>
> >> >>>>>
> >> >>>>>
> >> >>>>>
> >> >>>>> On Sat, Dec 10, 2011 at 12:07 AM, Lord Khan Han <
> >> >>> [email protected]
> >> >>>>> wrote:
> >> >>>>>
> >> >>>>>> I made major_compact but not flush...  will do now with flush..
> >> >>>>>>
> >> >>>>>>
> >> >>>>>>
> >> >>>>>> On Fri, Dec 9, 2011 at 11:58 PM, lars hofhansl <
> >> [email protected]
> >> >>>>> wrote:
> >> >>>>>>
> >> >>>>>>> Can you try flushing and compacting the table? How did you
> measure
> >> >>> the
> >> >>>>>>> size?
> >> >>>>>>>
> >> >>>>>>> Both can be done from the shell using the 'flush' and
> >> >> 'major_compact'
> >> >>>>>>> commands, resp.
> >> >>>>>>>
> >> >>>>>>>
> >> >>>>>>>
> >> >>>>>>> ----- Original Message -----
> >> >>>>>>> From: Lord Khan Han <[email protected]>
> >> >>>>>>> To: [email protected]
> >> >>>>>>> Cc:
> >> >>>>>>> Sent: Friday, December 9, 2011 1:50 PM
> >> >>>>>>> Subject: Hbase export / import Why doubling the Table Size ?
> >> >>>>>>>
> >> >>>>>>> Hi ,
> >> >>>>>>>
> >> >>>>>>> We are usng CDH3B4  and want to upgrade to CDH3u2.  Before doing
> >> >> this
> >> >>>>>>> we make a separate cluster with same config and installed
> CDH3u2.
> >> >>>>>>>
> >> >>>>>>> We exported our hbase table from cdh3b4  cluster  and import it
> to
> >> >>> the
> >> >>>>>>> new cdh3u2  cluster. Table is LZO and both cluster config is
> same.
> >> >>>>>>>
> >> >>>>>>> After import finished hbase table size doubled!! even its
> >> >> configured
> >> >>>>>>> to use LZO.  We changed table to snappy  import again and same
> >> >>> result.
> >> >>>>>>> Table size multiplied x 2  in new cdh3u2  cluster.
> >> >>>>>>>
> >> >>>>>>> We didnt find why ? Is there any ideas for this ?
> >> >>>>>>>
> >> >>>>>>> thanks
> >> >>>>>>>
> >> >>>>>>> Khan
> >> >>>>>>>
> >> >>>>>>>
> >> >>>>>>
> >> >>>>>
> >> >>>>
> >> >>>>
> >> >>>
> >> >>>
> >> >>
> >> >>
> >>
> >>
> >
>
>

Reply via email to