> -----Original Message----- > From: Mark Thomas <ma...@apache.org> > Sent: Wednesday, June 30, 2021 15:22 > To: users@tomcat.apache.org > Subject: Re: Possible Http11NioProtocol regression since 9.0.48? > > On 30/06/2021 13:57, André van der Lugt wrote: > > Hi, > > > > Since upgrading our Tomcat 9.0.x installation from 9.0.46 to 9.0.48, I've > noticed > > several times that the download of a response with static content from a > web archive > > (i.e. JavaScript file) takes 'keepAliveTimeout' time to complete. > > It only occurs with SSL enabled Http11NioProtocol and useSendfile=false. > > > > Could this be a bug? > > Possibly. > > > What is needed to further clarify this? > > A decrypted network trace (e.g. from wireshark) showing the problematic > connection. > > You'll need to set the SSLKEYLOGFILE environment variable. >
I finally managed to create a decrypted Wireshark capture with injected TLS session keys, will send it in a direct message due to size. I hope it provides the information needed. > How often does this occur? Is it on every request? The bootstrap client program is loading a number of Javascript files, with only one of them exceeding 700 KB compressed size. The first time after Tomcat start, everything is fine for all requests. When rerunning the bootstrap client program, the problem (taking 'keepAliveTimeout' to complete) occurs only for the 'big' Javascript file. And it stays like that with each next run. > > > The issue is clearly visible in the browser's developer tools on tab Timing: > > the Content Download phase takes its normal time to download the > content but > > also indicates "Caution: request is not finished yet!". > > The suggests some sort of issue with the client determining the content > length. > > Nothing in the changelog jumps out at me as a potential cause for this > at the moment. > > > > https://emea01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fto > mcat.apache.org%2Fmigration- > 9.html%23Upgrading_9.0.x&data=04%7C01%7C%7C1d1fc0d34df5420b5 > 79f08d93bca19e3%7C84df9e7fe9f640afb435aaaaaaaaaaaa%7C1%7C0%7C637 > 606561467284838%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAi > LCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C1000&sdata=cX > lU0Zg6Hwly7d4ZN0wUyfnstQ3UoNVnkdPCvwpvSmA%3D&reserved=0 > states this: > > "In 9.0.48 onwards, the NIO poller was simplified and the block poller > > and > > selector configuration has been removed" > > Could it be that the problem symptom is a consequence or side effect of > these changes? > > I wouldn't rule anything out at this point. > > Mark > > --------------------------------------------------------------------- > To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org > For additional commands, e-mail: users-h...@tomcat.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org For additional commands, e-mail: users-h...@tomcat.apache.org
