cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sylvain Lebresne (Commented) (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-3370) Deflate Compression corrupts SSTables
Date Tue, 18 Oct 2011 15:15:10 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-3370?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13129788#comment-13129788
] 

Sylvain Lebresne commented on CASSANDRA-3370:
---------------------------------------------

The exception is raised because there is a digest mismatch for the initial block of one of
the sstable.

Haven't been able to reproduce so far using stress with deflate with the default 1M keys (which
create a bunch of sstables, at least on my machine) using row slices and range scans (using
both Random and ByteOrdered partitioners).

Would you be able to 1) try with 1.0.0 and 2) try with the stress tool that comes with Cassandra
(it's in tools/stress of the source distribution, and you'll want to insert values with 'stress
-I DeflateCompressor' and read with 'stress -I DeflateCompressor -o RANGE_SLICE' ) and see
if you can reproduce? Another question is, did you used openJDK or Sun JDK?
                
> Deflate Compression corrupts SSTables
> -------------------------------------
>
>                 Key: CASSANDRA-3370
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-3370
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Core
>    Affects Versions: 1.0.0
>         Environment: Ubuntu Linux, amd64, Cassandra 1.0.0-rc2
>            Reporter: Christian Spriegel
>            Assignee: Sylvain Lebresne
>         Attachments: system.log
>
>
> Hi,
> it seems that the Deflate Compressor corrupts the SSTables. 3 out of 3 Installations
were corrupt. Snappy works fine.
> Here is what I did:
> 1. Start a single cassandra node (I was using ByteOrderedPartitioner)
> 2. Write data into cf that uses deflate compression - I think it has to be enough data
so that the data folder contains some files.
> 3. When I now try to read (I did a range scan) from my application, it fails and the
logs show corruptions:
> Caused by: org.apache.cassandra.io.compress.CorruptedBlockException: (/home/cspriegel/Development/cassandra1/data/Test/Response-h-2-Data.db):
corruption detected, chunk at 0 of length 65536.
> regards,
> Christian

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators: https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

Mime
View raw message