GGraziadei opened a new issue, #2131:
URL: https://github.com/apache/stormcrawler/issues/2131

   The `crawler-conf.yaml` generated by the archetypes leaves Storm's 
inter-executor batching at its default (`topology.producer.batch.size: 1`), so 
no batching happens between the spout, fetcher, parser, indexer and status 
updater executors.
   
   Storm 3.0 adds `topology.producer.batch.dynamic`: when enabled together with 
a `topology.producer.batch.size` greater than 1, the effective batch size 
adapts at runtime between 1 and the configured size with AIMD, staying small 
under light load to keep latency low and growing under sustained load to 
preserve throughput. That removes the usual objection to batching in a crawl 
topology, where traffic is bursty and latency to the status updater matters.
   
   Proposal: enable it in the archetype configurations with a conservative 
batch size below `topology.max.spout.pending`, and document the Storm topology 
settings that affect a crawl (`topology.message.timeout.secs`, 
`topology.max.spout.pending`, the producer batch settings and the flush 
interval), which are currently not covered in the configuration documentation.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to