Mar 8th, 2013, 05:23 AM
Spring Batch Integration - Chunking and Multi-VM
In the spring-batch-integration site, Chunking and Multi-VM section, there is a "TODO:failure modes" mentioned.
I had tried out the spring-batch-integration MessageChannelPartitionHandler and I am almost fine with the functionality.
The only and the main problem I felt was about the timeout scenario. But that too is explained to some extend in this thread MessageChannelPartitionHandler-and-tx-timeout
In most other cases, the master job will fail whenever there is a remote step failure and we can rectify the issue and restart the job.
Although I have not tried out the other remote chunking classes, I felt this could be the main issue.
So I am not able to identify what exactly is the TODO mentioned in the site. Is it the timeout situation itself or is it about fault tolerance to the full extend (like hadoop does)?
Mar 8th, 2013, 09:52 AM
To be honest, that is probably just an old documentation artifact. Most of the failure scenarios are handled via the underlying infrastructure (transactions and guaranteed delivery). Is there a specific scenario that you've identified that you have concerns about?
Mar 11th, 2013, 12:05 AM
Thank you for the response. I have not seen any specific scenario and I felt it good enough for me. But because of the TODO, I was wondering if I might hit a deadend (which I can't foresee now) somewhere down the line and was bit reluctant to proceed further. It would be great if you can update the docs, so that people won't not be turned away, like I was about to be.
Tags for this Thread