I created a partitioned table, partitioned daily. If I query the earlier
partitions, everything works. The later ones fail with error:

hive> select substr(user_name,1,1),count(*) from u_s_h_b where
dtpartition='2010-10-24' group by substr(user_name,1,1) ;
Total MapReduce jobs = 1
Launching Job 1 out of 1
Number of reduce tasks not specified. Estimated from input data size: 1
In order to change the average load for a reducer (in bytes):
  set hive.exec.reducers.bytes.per.reducer=<number>
In order to limit the maximum number of reducers:
  set hive.exec.reducers.max=<number>
In order to set a constant number of reducers:
  set mapred.reduce.tasks=<number>
java.lang.ArrayIndexOutOfBoundsException: 0
    at
org.apache.hadoop.mapred.FileInputFormat.identifyHosts(FileInputFormat.java:556)
    at
org.apache.hadoop.mapred.FileInputFormat.getSplitHosts(FileInputFormat.java:524)
    at
org.apache.hadoop.mapred.FileInputFormat.getSplits(FileInputFormat.java:235)
......snip.......
    at
sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
    at java.lang.reflect.Method.invoke(Method.java:597)
    at org.apache.hadoop.util.RunJar.main(RunJar.java:186)
Job Submission failed with exception
'java.lang.ArrayIndexOutOfBoundsException(0)'
FAILED: Execution Error, return code 1 from
org.apache.hadoop.hive.ql.exec.MapRedTask

It turns out that 2010-10-24 is 257 days from the very first partition in my
dataset (2010-01-09):

| date_sub('2010-10-24',interval 257 day) |
+-----------------------------------------+
| 2010-02-09                              |

That seems like an interesting coincidence. But try as I might, the Great
Googles will not show me a way to tune this, or even if it is tuneable, or
expected. Has anyone else run into a 256-partition limit in Hive? How do you
work around it? Why is that even the limit?! Shouldn't it be more like
32-bit maxint??!!

Thanks!

--
Tim Ellis
Riot Games

I have way more than 256 partitions per table. AFAIK, there is no partition limit.

>From your stack trace, you have some host name issue somewhere.
From: Time Less [mailto:[EMAIL PROTECTED]]
Sent: Tuesday, May 03, 2011 6:52 PM
To: [EMAIL PROTECTED]
Subject: Maximum Number of Hive Partitions = 256?

I created a partitioned table, partitioned daily. If I query the earlier partitions, everything works. The later ones fail with error:

hive> select substr(user_name,1,1),count(*) from u_s_h_b where dtpartition='2010-10-24' group by substr(user_name,1,1) ;
Total MapReduce jobs = 1
Launching Job 1 out of 1
Number of reduce tasks not specified. Estimated from input data size: 1
In order to change the average load for a reducer (in bytes):
  set hive.exec.reducers.bytes.per.reducer=<number>
In order to limit the maximum number of reducers:
  set hive.exec.reducers.max=<number>
In order to set a constant number of reducers:
  set mapred.reduce.tasks=<number>
java.lang.ArrayIndexOutOfBoundsException: 0
    at org.apache.hadoop.mapred.FileInputFormat.identifyHosts(FileInputFormat.java:556)
    at org.apache.hadoop.mapred.FileInputFormat.getSplitHosts(FileInputFormat.java:524)
    at org.apache.hadoop.mapred.FileInputFormat.getSplits(FileInputFormat.java:235)
......snip.......
    at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
    at java.lang.reflect.Method.invoke(Method.java:597)
    at org.apache.hadoop.util.RunJar.main(RunJar.java:186)
Job Submission failed with exception 'java.lang.ArrayIndexOutOfBoundsException(0)'
FAILED: Execution Error, return code 1 from org.apache.hadoop.hive.ql.exec.MapRedTask

It turns out that 2010-10-24 is 257 days from the very first partition in my dataset (2010-01-09):

| date_sub('2010-10-24',interval 257 day) |
+-----------------------------------------+
| 2010-02-09                              |

That seems like an interesting coincidence. But try as I might, the Great Googles will not show me a way to tune this, or even if it is tuneable, or expected. Has anyone else run into a 256-partition limit in Hive? How do you work around it? Why is that even the limit?! Shouldn't it be more like 32-bit maxint??!!

Thanks!

--
Tim Ellis
Riot Games

same here ... we have way more than 256 partitions in multiple tables. I am
sure the issue has something to do with an empty string passed to the substr
function. can you validate that the table has no null/empty string for
user_name or try running the query with len(user_name) > 1 (not sure about
query syntax) ?

On Tue, May 3, 2011 at 7:02 PM, Steven Wong <[EMAIL PROTECTED]> wrote:

> I have way more than 256 partitions per table. AFAIK, there is no partition
> limit.
>
>
>
> From your stack trace, you have some host name issue somewhere.
>
>
>
>
>
> *From:* Time Less [mailto:[EMAIL PROTECTED]]
> *Sent:* Tuesday, May 03, 2011 6:52 PM
> *To:* [EMAIL PROTECTED]
> *Subject:* Maximum Number of Hive Partitions = 256?
>
>
>
> I created a partitioned table, partitioned daily. If I query the earlier
> partitions, everything works. The later ones fail with error:
>
> hive> select substr(user_name,1,1),count(*) from u_s_h_b where
> dtpartition='2010-10-24' group by substr(user_name,1,1) ;
> Total MapReduce jobs = 1
> Launching Job 1 out of 1
> Number of reduce tasks not specified. Estimated from input data size: 1
> In order to change the average load for a reducer (in bytes):
>   set hive.exec.reducers.bytes.per.reducer=<number>
> In order to limit the maximum number of reducers:
>   set hive.exec.reducers.max=<number>
> In order to set a constant number of reducers:
>   set mapred.reduce.tasks=<number>
> java.lang.ArrayIndexOutOfBoundsException: 0
>     at
> org.apache.hadoop.mapred.FileInputFormat.identifyHosts(FileInputFormat.java:556)
>     at
> org.apache.hadoop.mapred.FileInputFormat.getSplitHosts(FileInputFormat.java:524)
>     at
> org.apache.hadoop.mapred.FileInputFormat.getSplits(FileInputFormat.java:235)
> ......snip.......
>     at
> sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:25)
>     at java.lang.reflect.Method.invoke(Method.java:597)
>     at org.apache.hadoop.util.RunJar.main(RunJar.java:186)
> Job Submission failed with exception
> 'java.lang.ArrayIndexOutOfBoundsException(0)'
> FAILED: Execution Error, return code 1 from
> org.apache.hadoop.hive.ql.exec.MapRedTask
>
> It turns out that 2010-10-24 is 257 days from the very first partition in
> my dataset (2010-01-09):
>
> | date_sub('2010-10-24',interval 257 day) |
> +-----------------------------------------+
> | 2010-02-09                              |
>
> That seems like an interesting coincidence. But try as I might, the Great
> Googles will not show me a way to tune this, or even if it is tuneable, or
> expected. Has anyone else run into a 256-partition limit in Hive? How do you
> work around it? Why is that even the limit?! Shouldn't it be more like
> 32-bit maxint??!!
>
> Thanks!
>
> --
> Tim Ellis
> Riot Games
>

> I am sure the issue has something to do with an empty string passed to the
> substr function.
We can rule out the substr() function. I get the same stack trace with any
query like:

hive> select <anyColumn> from ushb where dtpartition='2010-10-25' limit 10;

But this query succeeds:

hive> select * from ushb where dtpartition='2010-10-25' limit 10 ;

So SOMETHING about the data makes Hive (Hadoop?) unhappy. More specifically
something about trying to select a particular column from the data on some
days. I'm looking at the data to see if I can sort what it is.

I have way more than 256 partitions per table. AFAIK, there is no partition
> limit.
>
> From your stack trace, you have some host name issue somewhere.
>

I see why you'd think that from the stack trace, though I can't imagine why
it'd have a "host name issue somewhere." The partition create statements
have no hostname component. The query has no hostname component.

This is definitely a curious problem.

--
Tim Ellis
Riot Games

> This is definitely a curious problem.
>

It's data corruption. The file is tab-separated, so I created a quick Perl
pipe to print out the number of tabs on a given line:

-bash-3.2$ hadoop fs -cat
/user/hive/warehouse/ushb/2010-10-25/data-2010-10-25 | perl -pe
's/[^\t\n]//g' | perl -pe 's/\t/-/g' | sort | uniq -c

The STDOUT was slightly disturbing:

      1 --
1552318 -------

The STDERR moreso:

11/05/04 11:07:49 INFO hdfs.DFSClient: No node available for block:
blk_-1511269407958713809_10494
file=/user/hive/warehouse/ushb/2010-10-25/data-2010-10-25
11/05/04 11:07:49 INFO hdfs.DFSClient: Could not obtain block
blk_-1511269407958713809_10494 from any node: java.io.IOException: No live
nodes contain current block. Will get new block locations from namenode and
retry...
11/05/04 11:07:52 INFO hdfs.DFSClient: No node available for block:
blk_-1511269407958713809_10494
file=/user/hive/warehouse/ushb/2010-10-25/data-2010-10-25
11/05/04 11:07:52 INFO hdfs.DFSClient: Could not obtain block
blk_-1511269407958713809_10494 from any node: java.io.IOException: No live
nodes contain current block. Will get new block locations from namenode and
retry...
11/05/04 11:07:58 WARN hdfs.DFSClient: DFS Read: java.io.IOException: Could
not obtain block: blk_-1511269407958713809_10494
file=/user/hive/warehouse/ushb/2010-10-25/data-2010-10-25
    at
org.apache.hadoop.hdfs.DFSClient$DFSInputStream.chooseDataNode(DFSClient.java:1977)
    at
org.apache.hadoop.hdfs.DFSClient$DFSInputStream.blockSeekTo(DFSClient.java:1784)
    at
org.apache.hadoop.hdfs.DFSClient$DFSInputStream.read(DFSClient.java:1932)
    at java.io.DataInputStream.read(DataInputStream.java:83)
(...etc)
cat: Could not obtain block: blk_-1511269407958713809_10494
file=/user/hive/warehouse/ushb/2010-10-25/data-2010-10-25

--
Tim Ellis
Riot Games

> It turns out that 2010-10-24 is 257 days from the very first partition in
> my dataset (2010-01-09):
>
> | date_sub('2010-10-24',interval 257 day) |
> +-----------------------------------------+
> | 2010-02-09                              |
>

I just noticed 257 days is FEBRUARY 9th, not JANUARY 9th, as the above
shows. So there isn't even a 256ness of this problem in the first place. The
human brain tends to pay attention to beginnings and ends of strings,
ignoring the middle.

--
Tim Ellis
Riot Games