One way could be to handle the failure of a Pig job in a wrapper script. Or
use the Java APIs to be able to see if the job completed successfully.
On Thu, Oct 18, 2012 at 2:03 PM, Dmitriy Ryaboy <[EMAIL PROTECTED]> wrote:
> That's a Hadoop mapreduce feature, not a Pig feature, so that request
> should go there.
> Can't really do the _failure thing though, if you think about it --
> programs can fail by crashing, in which case they might not be able to
> write a file. Or maybe they are not crashing, but there is a problem
> talking to HDFS, still can't write the failure.
> On Thu, Oct 18, 2012 at 12:43 AM, Zebeljan, Nebojsa
> <[EMAIL PROTECTED]> wrote:
> > Hi,
> > Pig writes a 0 byte _SUCCESS file, when the STORE has been successfully
> done. Is there something like a _FAILURE file that indicates that the STORE
> > I need to determine with a another application (java), if the pig run
> has finished and if it was successful or not.
> > Any approach is welcome – Thanks for your help!
> > Regards,
> > Nebo