partitionpersist

[Weide Zhang <we...@gmail.com>]

Hi,

Can any one tell me when partitionpersist function is being used to update
state ? When it receives the list of tuples for status update , does it
ensure all the tuples in a batch is included when invoke the API ? Also,
how does the status update guarantees exact once semantic if there is
worker node failure ? Is it possible same batch that being replayed will
invoke status updater again ?

I have read the trident doc but still confused about the internals about
the partitionpersist and maybe persistAggregate function as well.  Not sure
if there are good documentations on internals and fail over handling.  Doc
is really lack on Trident side I have to say.

Weide