Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
HBase >> mail # user >> Region in Transition


Copy link to this message
-
Re: Region in Transition
Lukas,

  I am not sure I understand what you are saying there.  What I thought you
did was:

hadoop fs -mv /hbase /tmp/hbase.old

Restart HBase

You should now have a clean HBase

hadoop fs -mv /tmp/hbase.old/<table_name>

Once you have moved all of the tables

./bin/hbase hbck -fixMeta -fixAssignments

This should put your hbase back into a happy state, if it doesn't not you
had underlying issues that need resolving.

On Mon, Oct 15, 2012 at 12:09 PM, Lukáš Drbal <[EMAIL PROTECTED]> wrote:

> Wow, its perfect "hack" but what about "exists" rows?
> For example when i have saved new version from row in /hbase i can't
> replace it from /hbase.old
>
> Thanks for time
>
> Lukas
>
> 2012/10/15 Kevin O'dell <[EMAIL PROTECTED]>:
> > Lukas,
> >
> >   A little trick you can use is to just copy the table directories into
> > your new /hbase dir and then use:
> >
> > hbck -fixMeta -fixAssignments
> >
> > This will pull in the tables into your new empty META.  If once you do
> this
> > Regions get stuck in transition again, you have bigger problems.
> >
> > On Mon, Oct 15, 2012 at 11:48 AM, Lukáš Drbal <[EMAIL PROTECTED]>
> wrote:
> >
> >> Hi again.
> >>
> >> > Does FSCK come back clean?  Are those the regions showing in
> transition?
> >> >  We are not going to be able to get a clear idea of what to do next
> until
> >> > we gather some more data.  At this point running repairs could put
> your
> >> > data in jeopardy.
> >>
> >>
> >> Ye, this is in transition, but from hadoop fsck are all ok and
> >> consistent - reported only 14 undereplicated blocks from another
> >> directory (not /hbase).
> >>
> >> I must bring cluster back online, so i move /hbase on hdfs to
> >> /hbase.old and start clear cluster with new tables :(
> >>
> >> Can you give me please any suggestion how i can put data from old
> >> files to new tables?
> >>
> >> Thanks
> >>
> >> My today's night will be very long :(
> >>
> >> Lukas
> >>
> >> >
> >> > On Mon, Oct 15, 2012 at 10:58 AM, Lukáš Drbal <[EMAIL PROTECTED]>
> >> wrote:
> >> >
> >> >> Kevin,
> >> >>
> >> >> Its more than one hour and region are still in transition.
> >> >> Here is output from hbck -detailed http://pastebin.com/f9wLx9LU
> >> >>
> >> >> Thanks for time
> >> >>
> >> >> Lukas
> >> >>
> >> >> 2012/10/15 Kevin O'dell <[EMAIL PROTECTED]>:
> >> >> > Lukas,
> >> >> >
> >> >> >   Sure, how long has it been since the restart?  You will need to
> give
> >> >> the
> >> >> > regions time to transition and logs time to split.  Did the region
> >> >> > in transition properly?  Can you please out a pastebin together of
> >> hbck
> >> >> > -details full output so that I can take a look at it.  Once I have
> >> >> reviewed
> >> >> > it, I should be able to provide you with a plan to out this back
> >> >> together.
> >> >> >
> >> >> > On Mon, Oct 15, 2012 at 10:24 AM, Lukáš Drbal <
> [EMAIL PROTECTED]>
> >> >> wrote:
> >> >> >
> >> >> >> Hello Kevin,
> >> >> >>
> >> >> >> thanks for response. I clear it now and start cluster, but now i
> have
> >> >> >> many other's regions :(
> >> >> >> Here is output from hbase hbck -fixmeta
> >> >> >> http://pastebin.com/HqsPVLMi
> >> >> >>
> >> >> >> Any next hint?
> >> >> >>
> >> >> >> Thanks a lot
> >> >> >>
> >> >> >> 2012/10/15 Kevin O'dell <[EMAIL PROTECTED]>:
> >> >> >> > Have you tried clearing out your Znode information?  Typically,
> >> when I
> >> >> >> have
> >> >> >> > encountered a RIT, we will bring down HBase and go to the ZKcli
> and
> >> >> clear
> >> >> >> > out /hbase.  What do you see in the logs pertaining to the
> region?
> >> >>  If it
> >> >> >> > is a region that has bad hfiles or something like that you will
> >> not be
> >> >> >> able
> >> >> >> > to easily get rid of it through clearing out the ZK.
> >> >> >> >
> >> >> >> > On Mon, Oct 15, 2012 at 8:59 AM, Lukáš Drbal <
> >> [EMAIL PROTECTED]>
> >> >> >> wrote:
> >> >> >> >
> >> >> >> >> Oh sorry. My HBase version is 0.94.1 and hadoop 1.0.3
> >> >> >> >>
> >> >> >> >> I found this in master log http://pastebin.com/CYt5PZCL and

Kevin O'Dell
Customer Operations Engineer, Cloudera