fsparv commented on PR #3398: URL: https://github.com/apache/solr/pull/3398#issuecomment-3063102339
I also was unable to break the new code using the limit tests (100 runs each), I'm still curious if there's a reproducible method for producing the problematic behavior with the suspected deadlocks, and observing the +50ms delay. I suspect the delay may have been a memory consistency issue (perhaps causing poll() to not return despite an available item). I still haven't come up with a logic to understand any deadlock. I had previously been reluctant to change responseFutureMap to a ConcurrentHashMap since there's a different contract for nulls and it's pretty central, so any performance cost is a big deal (though it's usage is shard/request proportional, and not document/field/token proportional so perhaps I shouldn't worry so much?). Also, I worry that folks who have implemented custom handlers could be disrupted by changing the underlying types. What are your thoughts on those considerations? -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
