I have a custom processor that implements AbstractProcessor and a custom store 
that implements StateStore.

Before Kafka 1.0.0 the processors "init" method gets called after the state 
store is restored from changelog and that is good.
With Kafka 1.0.0 the processors "init" method is called BEFORE the state store 
is restored from changelog and that is bad.

My processor can only initialize when it has access to the state. However at 
the time KStreams calls "init" on the processor the state store may not have 
any data. It is not an option for me to initialize the processor lazily when a 
record arrives, or to re-initialize it when "onRestoreEnd" is called (its only 
called on restore; The state store "init" gets called before processor "init" 
regardless of restore or not.)

I think I need to have either of these:
a) know whether or not a state restore will take place and when not
b) or get a call to the state store regardless of whether state restore took 
place or not
c) or I need a "ready" method on the processor that gets called when the state 
store has completed restoring and is actually usable

Please help, thank you in advance.
Rainer

Reply via email to