Actually we do eat the exception, but we drop the component that caused(during creating/configuration) it (which is why the multiport syslog source was dropped). I am not exactly sure why the logging isn't detailed enough, but the only issue is that the exception isn't logged correctly. Again, this is something that has existed in Flume-1.2.0 also. I am not sure this is a release blocker.
-- Hari Shreedharan On Wednesday, November 28, 2012 at 2:17 PM, Brock Noland wrote: > Hmm, it looked to me like we were eating an exception. Maybe I am wrong. I'll > look at it tomorrow. > > -- > Brock Noland > Sent with Sparrow (http://www.sparrowmailapp.com/?sig) > > > On Wednesday, November 28, 2012 at 4:08 PM, Hari Shreedharan wrote: > > > Actually error reporting needs to be improved in the entire config system - > > which is a larger undertaking. I don't think the release needs to wait for > > that. Let's commit FLUME-1743 and then roll another (hopefully final) RC. > > > > Thanks, > > Hari > > > > -- > > Hari Shreedharan > > > > > > On Wednesday, November 28, 2012 at 1:58 PM, Brock Noland wrote: > > > > > I think 1743 should also make this obvious to the user. Our error > > > reporting here needs to be be improved. > > > > > > -- > > > Brock Noland > > > Sent with Sparrow (http://www.sparrowmailapp.com/?sig) > > > > > > > > > On Wednesday, November 28, 2012 at 3:45 PM, Will McQueen wrote: > > > > > > > Hi Brock, > > > > > > > > Initially I saw some issue with mismatched source file, but I no longer > > > > see > > > > that after doing a 'mvn clean install' instead of just 'mvn install'. > > > > Now, > > > > I see the same problems as Roshan and Alexander: in both trunk and > > > > flume-1.3.0, the multiport syslog tcp source is failing to configure > > > > regardless of whether we specify the FQCN or the short alias. > > > > > > > > With the patch that Hari submitted to FLUME-1743, I no longer see the > > > > issue > > > > in trunk and flume-1.3.0 branches. Given that there's neither the FQCN > > > > nor > > > > the alias work without the patch, I would consider this to be a blocker > > > > for > > > > 1.3.0 release so I've upgraded the FLUME-1743's priority to "blocker". > > > > > > > > Cheers, > > > > Will > > > > > > > > > > > > On Wed, Nov 28, 2012 at 12:59 PM, Brock Noland <[email protected] > > > > (mailto:[email protected])> wrote: > > > > > > > > > Hmm it looks like we are eating whatever exception is occurring in > > > > > FlumeConfiguration.valdiateSources and > > > > > then ComponentConfigurationFactory.create(). > > > > > > > > > > I'll look at this tomorrow unless someone else gets to it sooner. > > > > > > > > > > Brock > > > > > > > > > > On Wed, Nov 28, 2012 at 2:46 PM, Roshan Naik <[email protected] > > > > > (mailto:[email protected])> > > > > > wrote: > > > > > > > > > > > In rc5, both FQCN and short name gave same error. > > > > > > > > > > > > -roshan > > > > > > > > > > > > > > > > > > > > -- > > > > > Apache MRUnit - Unit testing MapReduce - > > > > > http://incubator.apache.org/mrunit/ > > > > > > > > > > > > > > > > > > > > > > > > > >
