BLOCKED "reactor-http-nio-*" threads under load [SPR-15874] #20429
Labels
in: web
Issues in web modules (web, webmvc, webflux, websocket)
status: duplicate
A duplicate of another issue
Eduard Dautov opened SPR-15874 and commented
I created POC of webflux stub server and faced with performance problems - my application can't handle more than 1000 users simultaneously.
!slow.png|thumbnail!
I made some investigations and found the root cause - all "reactor-HTTP" threads spent a lot of time in BLOCKED state.
!slowVM.png|thumbnail!
This happens because of org.springframework.http.codec.AbstractCodecConfigurer.AbstractDefaultCodecs#jackson2JsonEncoder() - if you don't specify Jackson2JsonEncoder, Configurer will create new Encoder.
In my application, I avoid this using WebFluxConfigurer
Results:
!fast.png|thumbnail!
!fastVM.png|thumbnail!
Source code of my application with Gatling tests https://github.com/Zajs/webflux-test
This bug can be related to #20143.
Affects: 5.0 RC3
Attachments:
Issue Links:
The text was updated successfully, but these errors were encountered: