hadoop-common-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jason Lowe (JIRA)" <j...@apache.org>
Subject [jira] [Updated] (HADOOP-14919) BZip2 drops records when reading data in splits
Date Wed, 04 Oct 2017 22:17:00 GMT

     [ https://issues.apache.org/jira/browse/HADOOP-14919?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

Jason Lowe updated HADOOP-14919:
    Attachment: HADOOP-14919.001.patch

A lot of the troubles with split handling in this codec are related to these issues:
1) It seeks _backwards_ from the split start looking for a possible bzip2 block header and
sometimes miscomputes where that should start.
2) It is reporting stream positions that can skip data that has not been processed yet (as
in this case).

So here's my pitch at fixing this for the Nth time:
- No seeking backwards.  Any block whose start marker appears before the split is not the
responsibility of this split's reader.  Split processing should never require reading data
just before the split offset, as that data is entirely the responsibility of the previous
split's reader.
- Stream position reports the byte where the block start marker begins rather than the byte
after it ends.  That way we're always consistent about who is responsible for "consuming"
a block start header and never report byte positions that may have skipped some data bits.

This is probably going to break _something_ given how many attempts there have been at fixing
this, so I greatly appreciate any and all eyes willing to take a look.  Alternative proposals
are also welcome.

Attaching a patch that implements the approach described above.  Thanks to [~tanakahda] for
the unit test.  I extended it to test all the split positions around the block start marker
for this test case. Speaking of tests, Jenkins won't run all the related tests, so I also
manually ran the TestLineRecordReader tests in hadoop-mapreduce-client-core and they passed.

> BZip2 drops records when reading data in splits
> -----------------------------------------------
>                 Key: HADOOP-14919
>                 URL: https://issues.apache.org/jira/browse/HADOOP-14919
>             Project: Hadoop Common
>          Issue Type: Bug
>    Affects Versions: 2.8.0, 2.7.3, 3.0.0-alpha1
>            Reporter: Aki Tanaka
>            Assignee: Jason Lowe
>            Priority: Critical
>         Attachments: 250000.bz2, HADOOP-14919.001.patch, HADOOP-14919-test.patch
> BZip2 can drop records when reading data in splits. This problem was already discussed
before in HADOOP-11445 and HADOOP-13270. But we still have a problem in corner case, causing
lost data blocks.
> I attached a unit test for this issue. You can reproduce the problem if you run the unit
> First, this issue happens when position of newly created stream is equal to start of
split. Hadoop has some test cases for this (blockEndingInCR.txt.bz2 file for TestLineRecordReader#testBzip2SplitStartAtBlockMarker,
etc). However, the issue I am reporting does not happen when we run these tests because this
issue happens only when the start of split byte block includes both block marker and compressed
> BZip2 block marker - 0x314159265359 (001100010100000101011001001001100101001101011001)
> blockEndingInCR.txt.bz2 (Start of Split - 136504):
> {code:java}
> $ xxd -l 6 -g 1 -b -seek 136498 ./hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-core/target/test-classes/blockEndingInCR.txt.bz2
> 0021532: 00110001 01000001 01011001 00100110 01010011 01011001  1AY&SY
> {code}
> Test bz2 File (Start of Split - 203426)
> {code:java}
> $ xxd -l 7 -g 1 -b -seek 203419 250000.bz2
> 0031a9b: 11100110 00101000 00101011 00100100 11001010 01101011  .(+$.k
> 0031aa1: 00101111                                               /
> {code}
> Let's say a job splits this test bz2 file into two splits at the start of split (position
> The former split does not read records which start position 203426 because BZip2 says
the position of these dropped records is 203427. The latter split does not read the records
because BZip2CompressionInputStream read the block from position 320955.
> Due to this behavior, records between 203427 and 320955 are lost.
> Also, if we reverted the changes in HADOOP-13270, we will not see this issue. We will
see HADOOP-13270 issue though.

This message was sent by Atlassian JIRA

To unsubscribe, e-mail: common-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: common-issues-help@hadoop.apache.org

View raw message