Hi,

I recently tried to use sqoop to export a Hive table that lives on S3 into my MySQL server (sqoop export --options-file config.txt --table _universe --export-dir s3n://key:secret@mybucket/universe --input-fields-terminated-by '\0001' -m 1 --input-null-string '\\N' --input-null-non-string '\\N'^C). My Sqoop runs on a CDH4 cluster on EC2. I was getting errors such as the following:

13/02/11 17:37:15 ERROR security.UserGroupInformation: PriviledgedActionException as:XXX (auth:SIMPLE) cause:java.io.FileNotFoundException: File does not exist: /universe/000000_0.snappy
13/02/11 17:37:15 ERROR tool.ExportTool: Encountered IOException running export job: java.io.FileNotFoundException: File does not exist: /universe/000000_0.snappy

Since the files do exist on S3, I was reminded of getting the same errors when running Hive queries against this table. The reason Hive was failing back then is because of a bug in CombineFileInputFormat when using it against a non-default file system. These issues have since been fixed in Hadoop:
https://issues.apache.org/jira/browse/MAPREDUCE-1806
https://issues.apache.org/jira/browse/MAPREDUCE-2704
I believe Sqoop uses a version of CombineFileInputFormat but as far as I can tell from the latest sources on GIT hasn't incorporated the above fixes. My questions for the user group:
Am I completely off in my investigations?
Is there something I am missing in configuring Sqoop for exporting from S3?
Is there a way for me to bypass the CombineFileInputFormat so I can make my exports work?

Many thanks,

Jurgen

Hi Jurgen,
I believe that your investigation is going very good direction, the MAPREDUCE-1806 in deed do not seem to be ported into Sqoop. Would you mind opening a JIRA for that [1]?

Jarcec

Links:
1: https://issues.apache.org/jira/browse/SQOOP

On Mon, Feb 11, 2013 at 09:01:01PM +0000, Jurgen Van Gael wrote:
> Hi,
>
> I recently tried to use sqoop to export a Hive table that lives on S3 into my MySQL server (sqoop export --options-file config.txt --table _universe --export-dir s3n://key:secret@mybucket/universe --input-fields-terminated-by '\0001' -m 1 --input-null-string '\\N' --input-null-non-string '\\N'^C). My Sqoop runs on a CDH4 cluster on EC2. I was getting errors such as the following:
>
> 13/02/11 17:37:15 ERROR security.UserGroupInformation: PriviledgedActionException as:XXX (auth:SIMPLE) cause:java.io.FileNotFoundException: File does not exist: /universe/000000_0.snappy
> 13/02/11 17:37:15 ERROR tool.ExportTool: Encountered IOException running export job: java.io.FileNotFoundException: File does not exist: /universe/000000_0.snappy
>
> Since the files do exist on S3, I was reminded of getting the same errors when running Hive queries against this table. The reason Hive was failing back then is because of a bug in CombineFileInputFormat when using it against a non-default file system. These issues have since been fixed in Hadoop:
> https://issues.apache.org/jira/browse/MAPREDUCE-1806
> https://issues.apache.org/jira/browse/MAPREDUCE-2704
>
>
> I believe Sqoop uses a version of CombineFileInputFormat but as far as I can tell from the latest sources on GIT hasn't incorporated the above fixes. My questions for the user group:
> Am I completely off in my investigations?
> Is there something I am missing in configuring Sqoop for exporting from S3?
> Is there a way for me to bypass the CombineFileInputFormat so I can make my exports work?
>
> Many thanks,
>
> Jurgen
>

Sure thing. Just added it as SQOOP-891 [1], let me know if there are things I need to change.

Jurgen

Links:
1: https://issues.apache.org/jira/browse/SQOOP-891

On Tuesday, 12 February 2013 at 16:12, Jarek Jarcec Cecho wrote:

> Hi Jurgen,
> I believe that your investigation is going very good direction, the MAPREDUCE-1806 in deed do not seem to be ported into Sqoop. Would you mind opening a JIRA for that [1]?
>
> Jarcec
>
> Links:
> 1: https://issues.apache.org/jira/browse/SQOOP
>
> On Mon, Feb 11, 2013 at 09:01:01PM +0000, Jurgen Van Gael wrote:
> > Hi,
> >
> > I recently tried to use sqoop to export a Hive table that lives on S3 into my MySQL server (sqoop export --options-file config.txt --table _universe --export-dir s3n://key:secret@mybucket/universe --input-fields-terminated-by '\0001' -m 1 --input-null-string '\\N' --input-null-non-string '\\N'^C). My Sqoop runs on a CDH4 cluster on EC2. I was getting errors such as the following:
> >
> > 13/02/11 17:37:15 ERROR security.UserGroupInformation: PriviledgedActionException as:XXX (auth:SIMPLE) cause:java.io.FileNotFoundException: File does not exist: /universe/000000_0.snappy
> > 13/02/11 17:37:15 ERROR tool.ExportTool: Encountered IOException running export job: java.io.FileNotFoundException: File does not exist: /universe/000000_0.snappy
> >
> > Since the files do exist on S3, I was reminded of getting the same errors when running Hive queries against this table. The reason Hive was failing back then is because of a bug in CombineFileInputFormat when using it against a non-default file system. These issues have since been fixed in Hadoop:
> > https://issues.apache.org/jira/browse/MAPREDUCE-1806
> > https://issues.apache.org/jira/browse/MAPREDUCE-2704
> >
> >
> > I believe Sqoop uses a version of CombineFileInputFormat but as far as I can tell from the latest sources on GIT hasn't incorporated the above fixes. My questions for the user group:
> > Am I completely off in my investigations?
> > Is there something I am missing in configuring Sqoop for exporting from S3?
> > Is there a way for me to bypass the CombineFileInputFormat so I can make my exports work?
> >
> > Many thanks,
> >
> > Jurgen