Storm failover, is it possible to know the last succeeded bolt when fail method gets called in spout?

[Xueming Li <ja...@gmail.com>]

Hi guys,

For a Storm topology, if a tuple failed to be processed (either timed out
or fail method got called on the tuple explicitly), I understand that fail
method of the spout will be called with the tuple message id. Other than
that, can Storm provide any info as to the last bolt that had processed the
tuple successfully? Thanks in advance!

Thanks,
James

Re: Storm failover, is it possible to know the last succeeded bolt when fail method gets called in spout?

[Xueming Li <ja...@gmail.com>]

Thank you very much Taylor for prompt reply :-)

Re: Storm failover, is it possible to know the last succeeded bolt when fail method gets called in spout?

["P. Taylor Goetz" <pt...@gmail.com>]

In short, no. Not today.

Storm's acking mechanism is largely stateless (one long can track an entire tuple tree),which is one of the reasons it is so efficient.

But the acking mechanism is also based on Storms core primitives, so it is entirely possible.

There is a JIRA for adding additional metadata to the spout fail method (specific number escapes me ATM). We should consider adding this as as a related issue.

-Taylor


> On May 27, 2014, at 8:58 PM, Xueming Li <ja...@gmail.com> wrote:
> 
> Hi guys,
> 
> For a Storm topology, if a tuple failed to be processed (either timed out or fail method got called on the tuple explicitly), I understand that fail method of the spout will be called with the tuple message id. Other than that, can Storm provide any info as to the last bolt that had processed the tuple successfully? Thanks in advance!
> 
> Thanks,
> James