> On Feb. 3, 2014, 4:58 p.m., Jarek Cecho wrote: > > Hi Venkat, > > thank you very much for working on this one! Allowing user to set the > > transaction isolation level seems to be good idea. I'm a bit concerned that > > the change is only in the DBInputFormat and as a result it won't be > > utilized by connectors using their own input formats, but I guess that > > there is not much we can do about that in Sqoop 1. Hopefully we will have > > better situation in Sqoop 2! Couple of high level notes: > > > > 1) Do you think that it would be useful to let user set any transaction > > level? Right now Sqoop allows only two, so perhaps allowing all of them > > might be useful? > > 2) Do you think that it would be useful to set the transaction isolation > > also for connections created from within the Sqoop process (e.g. while > > fetching metadata, ...). > > 3) Would you mind documenting the new option in the user guide? > > > > Jarcec > > Venkat Ranganathan wrote: > Good questions Jarcec. Thanks for the quick review also. > > 1) While it may seem that it may seem good to allow all isolation > levels, the other isolation levels (repeatable read and serializable) have > increased transaction isolation requirements and may not be really useful > 2) Metadata queries can be read committed (internally some DBMSs do use > recursive queries for them) and show that should be fine. For those queries > using the connection to get the result set metadata (like 1=0 clause query) > do not end up in acquiring any locks so should be OK as it is > 3) Will definitely add doc. Sorry for missing it >
Thank you for the answers Venkat, makes complete sense to me. - Jarek ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/17650/#review33434 ----------------------------------------------------------- On Feb. 3, 2014, 8:01 p.m., Venkat Ranganathan wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/17650/ > ----------------------------------------------------------- > > (Updated Feb. 3, 2014, 8:01 p.m.) > > > Review request for Sqoop. > > > Bugs: SQOOP-1278 > https://issues.apache.org/jira/browse/SQOOP-1278 > > > Repository: sqoop-trunk > > > Description > ------- > > Allow the use of read uncommitted isolation on all databases that support it. > Oracle does not read uncommitted as a settable option in the JDBC drivers so > it will behave as before on Oracle. > > > Diffs > ----- > > src/docs/user/common-args.txt 8a017f4 > src/docs/user/import.txt 0db6d97 > src/java/org/apache/sqoop/SqoopOptions.java 46e158c > src/java/org/apache/sqoop/mapreduce/DataDrivenImportJob.java b21560e > src/java/org/apache/sqoop/mapreduce/db/DBConfiguration.java be942ce > src/java/org/apache/sqoop/mapreduce/db/DBInputFormat.java 73ed94e > src/java/org/apache/sqoop/tool/BaseSqoopTool.java 6d6f1ea > src/test/com/cloudera/sqoop/TestSqoopOptions.java 686d398 > > Diff: https://reviews.apache.org/r/17650/diff/ > > > Testing > ------- > > One new test to test the option. Explicitly tested with Oracle, DB2 and > Postgresql. All unit tests passed > > > Thanks, > > Venkat Ranganathan > >
