kotman12 commented on PR #4974: URL: https://github.com/apache/solr/pull/4974#issuecomment-6092786097
> I didn't have a specific approach in mind to make it adjunct; it wasn't an entirely thought-out idea. > > This issue is related to #3428 which should probably be reverted now. If it gets reverted, then that provides a way for a client who exclusively uses the async API to abort the requests it send without aborting the entire client. Yes I did see that issue when developing this. It's a bit frustrating to have to work around http/2 issues (again). Seems that this stream reset circuit breaker is really only a problem with http/2. Claude had an idea to schedule the aborts on future cancel with a small delay so most completed before sending the reset to the rest (I think the issue in the linked PR was that there are too many resets too close in time). I actually almost went the async route with this PR. That being said I think its more natural to leave IF sync and I see no reason to not have cancelability on sync calls as well. I initially reached for the async API only because it _seemed_ like the intended cancellable mode the http jetty solr client supported (until I found cancellation wasn't propagated due to the above 😬). But it required a conversion layer to sync which was quite ugly. I support undoing the change you cite though I don't think it is directly related to this change anymore. Also, I dont want to introduce the reset overload issue again so we'd have to see what works regarding that. -- 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]
