why this _acker bolt is running on storm topology

[Somnath]

Hi All,

I am trying to run one topology with 1 spout and 3 bolt.

my topology looks like this .

spout name : getInput
bolts name : classBolt,keywordsbolt,Timeintervalbolt

and getInput spout is connected to classBolt which is connected to
keywordsbolt and Timeintervalbolt

while running the topology one more bolt _ackerbolt is running which
is acking allmost all records.

can anyone please explain why this extra ackerbolt is running and is
it good to let it run and acked tuples.

also i am seeing latency of around 390 ms in my spouts , what should i
do to less the latency period

i am using storm 0.6.0 and i am attaching my topolgy summary with this mail.

[Adrian Petrescu]

That's the bolt that implicitly takes care of Storm's message processing guarantee. It's the one that receives ack's from bolts, figures out when the entire tree has been processed, and sends the final ack (or fail) message back to the originating spout. You can get rid of it by setting numAckers to 0 in the topology config, but then you won't be able to ack or fail tuples if you do so.

You can read all about it here: https://github.com/nathanmarz/storm/wiki/Guaranteeing-message-processing

Cheers,

Adrian