Oozie-1972 is created for this issue: [OOZIE-1972] Oozie doesn't support unicode in appPath name for sql server - ASF JIRA
[OOZIE-1972] Oozie doesn't support unicode in appPath name for sql server - ASF JIRA By default, openJPA maps String to varchar(255). However, in sql server, varchar(255) doesn't support unicode encoding. View on issues.apache.org Preview by Yahoo ________________________________ From: Alejandro Abdelnur <[email protected]> To: bowen zhang <[email protected]> Cc: "[email protected]" <[email protected]> Sent: Wednesday, August 13, 2014 9:27 AM Subject: Re: oozie doesn't support unicode in app_path How this would affect an existing Oozie setup on sqlserver where the DB was created prior to this fix? Would that require a migration handling? I think it is time to file a JIRA and put all these findings there. ths On Tue, Aug 12, 2014 at 5:49 PM, bowen zhang <[email protected]> wrote: > One possible solution can be adding properties in oozie-sqlserver-orm.xml > to override the mapping from string to varchar. Instead, we want sqlserver > to map String to nvarchar. > > > On Tuesday, August 12, 2014 3:51 PM, bowen zhang < > [email protected]> wrote: > > > I think this is a doc issue when it comes to DB like mysql. But, it is a > real issue for sql server which oozie also supports. In sql server, varchar > or text don't support unicode and by default, openJPA mapping tool maps > "String" into "varchar(255)" instead of "nvarchar" or "ntext". Is there a > way for openJPA to not map "String" into "Varchar" during table creation? > If not, we need to do some specila handling for sql server in oozieDBCLI. > > > > On Tuesday, August 12, 2014 6:39 AM, Alejandro Abdelnur < > [email protected]> wrote: > > > So, it seems it is a JVM and DB configuration issue, right? It would be > great to verify this is the case and add the necessary config steps to the > documentation. > > thx > > > On Mon, Aug 11, 2014 at 2:03 PM, bowen zhang < > [email protected]> wrote: > > I think I figured out this issue. For oozie to support unicode in path > name, the JAVA_TOOL_OPTIONS in the JVM needs to use an encoding method that > supports unicode. And the oozie DB needs to use an encoding method that's > consistent with the JVM. By default, mysql, for example, uses "latin1" if > running "create database oozie;". This won't work since "latin1" only > supports Western characters. > > > > On Friday, August 8, 2014 7:06 PM, bowen zhang <[email protected]> > wrote: > > > > here is an example: > cat job.properties: > nameNode=hdfs://localhost:9000 > jobTracker=localhost:8032 > queueName=default > examplesRoot=examples > > oozie.wf.application.path=${nameNode}/user/${user.name > }/${examplesRoot}/apps/map-reduce文 > outputDir=map-reduce > > > > On Friday, August 8, 2014 6:55 PM, Alejandro Abdelnur <[email protected]> > wrote: > > > > hey bowen, it would help to know exactly where do you see issues, a simple > wf showing the issue would be great > > Alejandro > (phone typing) > > > > On Aug 8, 2014, at 18:12, bowen zhang <[email protected]> > wrote: > > > > Hi guys, > > I recently discovered a hole in oozie implementation where oozie breaks > when having hdfs app_path with unicode. If my app_path contains non-ascii > characters like Chinese words, the DB and server side will not retrieve it > correctly. Do we have strict utf-8 encoding throughout oozie? Any thoughts > on this? > > Bowen > > > > > -- > Alejandro > > > > > -- Alejandro
