You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@hbase.apache.org by Manuel de Ferran <ma...@gmail.com> on 2012/02/29 11:04:56 UTC
Upgrading to HBase-0.92
Greetings,
we're planning to move from HBase-0.90.3/Hadoop-0.20-append to
HBase-0.92/Hadoop-1.0.0. We did a couple of tests and we have a few
questions :
on first start, HBase master could not start, we had to rmr /hbase from
zookeeper. Is it safe ? By the way, why do we have to do that ?
Secondly, we're a bit worried about some old naming regions from an old
HBase 0.20 : is there any specific procedure for them ?
Lastly, we faced a weird "nil class" under hbase shell when we tried to
'list'. It seems that the root cause was oldlogfiles.log like
/hbase/<table_name>/227772165/oldlogfile.log. There was nothing but this
file under /hbase/<table_name>/227772165. Is it safe to remove them ?
Regards
Re: Upgrading to HBase-0.92
Posted by Manuel de Ferran <ma...@gmail.com>.
>
> > I could not find any reference to this region in status 'detailed'. And I
> > guess that if the region was in use, there should be additional files
> > under 196843901/ ?
>
> It seems that not only is that region isn't used, but that table isn't
> in use either? Is this right? If so you may have hit this bug:
> https://issues.apache.org/jira/browse/HBASE-5415
>
>
That's right. That's not a big deal we will clean up the HDFS from thoses
files.
Thanks
Re: Upgrading to HBase-0.92
Posted by Jean-Daniel Cryans <jd...@apache.org>.
On Thu, Mar 1, 2012 at 5:20 AM, Manuel de Ferran
<ma...@gmail.com> wrote:
> Well, I can not reproduce it. But I take the opportunity to digress a bit.
> Considering that this is a fresh start and that everything have been
> stopped gracefully : is there any risk to remove the zookeeper content ?
> What are the valuable data there ?
Nothing special if you're not using replication (in which case you
would have the replication metadata).
> Yes right, here are the details :
>
> hbase shell output :
>
> HBase Shell; enter 'help<RETURN>' for list of supported commands.
> Type "exit<RETURN>" to leave the HBase Shell
> Version 0.92.0, r1231986, Mon Jan 16 13:16:35 UTC 2012
>
> hbase(main):001:0> list
> TABLE
>
>
> ERROR: undefined method `map' for nil:NilClass
>
> Here is some help for this command:
> List all tables in hbase. Optional regular expression parameter could
> be used to filter the output. Examples:
>
> hbase> list
> hbase> list 'abc.*'
>
>
> gives the corresponding error in master logs :
> 2012-03-01 08:56:33,338 WARN org.apache.hadoop.hbase.master.HMaster: Failed
> getting all descriptors
> org.apache.hadoop.hbase.TableExistsException: No descriptor for gre014749
> walking thru the HDFS:
> hadoop_storage fs -lsr /hbase/gre014749
> drwxr-xr-x - hadoop supergroup 0 2011-07-22 07:24
> /hbase/gre014749/196843901
> -rw-r--r-- 3 hadoop supergroup 188392 2011-07-22 07:24
> /hbase/gre014749/196843901/oldlogfile.log
>
> I could not find any reference to this region in status 'detailed'. And I
> guess that if the region was in use, there should be additional files
> under 196843901/ ?
It seems that not only is that region isn't used, but that table isn't
in use either? Is this right? If so you may have hit this bug:
https://issues.apache.org/jira/browse/HBASE-5415
J-D
Re: Upgrading to HBase-0.92
Posted by Manuel de Ferran <ma...@gmail.com>.
Thanks for your reply J-D
my comments, inline
> > on first start, HBase master could not start, we had to rmr /hbase from
> > zookeeper. Is it safe ? By the way, why do we have to do that ?
>
> You should not have to do that, what was the reason the master didn't
> start? It's hard to answer "By the way, why do we have to do that ?"
> if you don't give us anything to ponder about.
>
>
Well, I can not reproduce it. But I take the opportunity to digress a bit.
Considering that this is a fresh start and that everything have been
stopped gracefully : is there any risk to remove the zookeeper content ?
What are the valuable data there ?
> > Lastly, we faced a weird "nil class" under hbase shell when we tried to
> > 'list'. It seems that the root cause was oldlogfiles.log like
> > /hbase/<table_name>/227772165/oldlogfile.log. There was nothing but this
> > file under /hbase/<table_name>/227772165. Is it safe to remove them ?
>
> Woah hold on there, there's a lot of info missing.
>
> You say "we faced a weird 'nil class' ", what was it like?
>
> Then "It seems that the root cause was oldlogfiles.log", how so?
>
> And about "Is it safe to remove them ?", it depends... is 227772165 an
> active region? Is the file old? If yes to both, then it's safe to
> delete as it would just be an old leftover which was fixed in the
> meantime.
>
>
Yes right, here are the details :
hbase shell output :
HBase Shell; enter 'help<RETURN>' for list of supported commands.
Type "exit<RETURN>" to leave the HBase Shell
Version 0.92.0, r1231986, Mon Jan 16 13:16:35 UTC 2012
hbase(main):001:0> list
TABLE
ERROR: undefined method `map' for nil:NilClass
Here is some help for this command:
List all tables in hbase. Optional regular expression parameter could
be used to filter the output. Examples:
hbase> list
hbase> list 'abc.*'
gives the corresponding error in master logs :
2012-03-01 08:56:33,338 WARN org.apache.hadoop.hbase.master.HMaster: Failed
getting all descriptors
org.apache.hadoop.hbase.TableExistsException: No descriptor for gre014749
at
org.apache.hadoop.hbase.util.FSTableDescriptors.get(FSTableDescriptors.java:164)
at
org.apache.hadoop.hbase.util.FSTableDescriptors.getAll(FSTableDescriptors.java:182)
at
org.apache.hadoop.hbase.master.HMaster.getHTableDescriptors(HMaster.java:1554)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at
sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:39)
at
sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
at java.lang.reflect.Method.invoke(Method.java:597)
at
org.apache.hadoop.hbase.ipc.WritableRpcEngine$Server.call(WritableRpcEngine.java:364)
at
org.apache.hadoop.hbase.ipc.HBaseServer$Handler.run(HBaseServer.java:1326)
walking thru the HDFS:
hadoop_storage fs -lsr /hbase/gre014749
drwxr-xr-x - hadoop supergroup 0 2011-07-22 07:24
/hbase/gre014749/196843901
-rw-r--r-- 3 hadoop supergroup 188392 2011-07-22 07:24
/hbase/gre014749/196843901/oldlogfile.log
I could not find any reference to this region in status 'detailed'. And I
guess that if the region was in use, there should be additional files
under 196843901/ ?
Re: Upgrading to HBase-0.92
Posted by Jean-Daniel Cryans <jd...@apache.org>.
Inline.
J-D
On Wed, Feb 29, 2012 at 2:04 AM, Manuel de Ferran
<ma...@gmail.com> wrote:
> Greetings,
>
> we're planning to move from HBase-0.90.3/Hadoop-0.20-append to
> HBase-0.92/Hadoop-1.0.0. We did a couple of tests and we have a few
> questions :
>
> on first start, HBase master could not start, we had to rmr /hbase from
> zookeeper. Is it safe ? By the way, why do we have to do that ?
You should not have to do that, what was the reason the master didn't
start? It's hard to answer "By the way, why do we have to do that ?"
if you don't give us anything to ponder about.
> Secondly, we're a bit worried about some old naming regions from an old
> HBase 0.20 : is there any specific procedure for them ?
You don't have to be worried and if you really want to get rid of them
AFAIK the only way would be to split them or copy them to another
table.
> Lastly, we faced a weird "nil class" under hbase shell when we tried to
> 'list'. It seems that the root cause was oldlogfiles.log like
> /hbase/<table_name>/227772165/oldlogfile.log. There was nothing but this
> file under /hbase/<table_name>/227772165. Is it safe to remove them ?
Woah hold on there, there's a lot of info missing.
You say "we faced a weird 'nil class' ", what was it like?
Then "It seems that the root cause was oldlogfiles.log", how so?
And about "Is it safe to remove them ?", it depends... is 227772165 an
active region? Is the file old? If yes to both, then it's safe to
delete as it would just be an old leftover which was fixed in the
meantime.
J-D