----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/67675/#review205494 -----------------------------------------------------------
Hi Fero, If I understand correclty, with this patch we're only displaying a warning when using --resilient to let the users know they should add --split-by (even if they do so?). In the documentation you're saying omitting --split-by can lead to lost/duplicated records. Shouldn't we stop the importing if there's no --split-by then? I understand we can't enforce the uniqeness and ascending order though, so keeping some kind of warning could make sense too. What do you think? Regards, Daniel - daniel voros On June 25, 2018, 3:17 p.m., Fero Szabo wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/67675/ > ----------------------------------------------------------- > > (Updated June 25, 2018, 3:17 p.m.) > > > Review request for Sqoop, Boglarka Egyed, daniel voros, and Szabolcs Vasas. > > > Bugs: SQOOP-3332 > https://issues.apache.org/jira/browse/SQOOP-3332 > > > Repository: sqoop-trunk > > > Description > ------- > > This is the documentation part of SQOOP-3333. > > > Diffs > ----- > > src/docs/user/connectors.txt f1c7aebe > src/java/org/apache/sqoop/manager/SQLServerManager.java c98ad2db > src/java/org/apache/sqoop/manager/SqlServerManagerContextConfigurator.java > cf58f631 > src/test/org/apache/sqoop/manager/sqlserver/SQLServerManagerImportTest.java > fc1c4895 > > > Diff: https://reviews.apache.org/r/67675/diff/2/ > > > Testing > ------- > > Unit tests, 3rdparty tests, ant docs. > > I've also investigated how export and import works: > > Import has it's retry mechanism in > org.apache.sqoop.mapreduce.db.SQLServerDBRecordReader#nextKeyValue > In case of error, it re-calculates the db query, thus the implicit > requirements > > Export has it's retry loop in > org.apache.sqoop.mapreduce.SQLServerAsyncDBExecThread#write > It doesn't recalculate the query, thus is a lot safer. > > > Thanks, > > Fero Szabo > >
