java - storm + kafka: understanding ack, fail and latency -


i'm using kafkaspout consume 2 kafka topics each of has 6 partitions. spout goes single bolt unpack relevant bytes , second bolt further processing.

when @ storm-ui numbers aren't making sense , i'm hoping can shed light.

  1. the kafka spout says 'acked' ~3600 tuples , failed ~73m. looking @ bolts in next group see have acked ~73m 0 failed while other have acked ~1.3m (no fails). shouldn't these numbers line somehow?

  2. the 'complete latency' in row spout ~2500ms while execute , process latency bolts ranges <1ms ~50ms. again - what's correspondence?

yes - topology has major issues (see this related question).

i'm attaching image of ui in hopes of helping me understand it.

enter image description here

it sounds tuples timing out , spout being notified. 0 fails reported bolts means bolts didn't explicitly fail tuples, why think must have timed out.

per other question, because bolts getting work, you're getting severe backlogs cause of tuple timeouts.

per comment in other question, can change shuffle grouping spread workload bolts , increase timeout setting in config.


Comments

Popular posts from this blog

c# - Validate object ID from GET to POST -

node.js - Custom Model Validator SailsJS -

php - Find a regex to take part of Email -