Re: Upgrade to Primefaces 6.2, Primeface Extensions6.2.2, Atlantis 1.0.3
Posted: 08 May 2018, 15:44
We started to get similar log messages one or two months ago without upgrading PrimeFaces, using Wildfly 10.1.0. The post in the Wildfly/JBoss mailing list describes an issue with HTTPS or HTTP/2, but we only use HTTP 1.1. Also, in our case, the application works as usual - the only annoyance is these log messages about broken pipes.
I also use Firefox and looked into the actual TCP requests via Wireshark. The log messages correspond to HTTP requests that are canceled immediately after submitting them, before the server could send the response. I strongly suspect the "Race Cache With Network" feature which was enabled in Firefox 59 (released March 13) to be the cause.
With this feature, Firefox _may_ send requests for cached resources concurrently to the local disk cache and the actual network server. Whichever result arrives first is used - the other request is canceled (which would cause a log message about a "Broken pipe", as with other connection aborts).
The feature can be disabled in Firefox in about:config by setting network.http.rcwn.enabled to false. For me, I stopped getting log messages after setting this to false. Of course this is not a solution, but the log messages seem to be benign in our case and could be ignored.
For more information about this Firefox feature, see https://slides.com/valentingosu/race-ca ... twork-2017
I believe a future version of Firefox might also mark these "racing" requests in the network panel, which would make identifying problems like this a lot easier.
I also use Firefox and looked into the actual TCP requests via Wireshark. The log messages correspond to HTTP requests that are canceled immediately after submitting them, before the server could send the response. I strongly suspect the "Race Cache With Network" feature which was enabled in Firefox 59 (released March 13) to be the cause.
With this feature, Firefox _may_ send requests for cached resources concurrently to the local disk cache and the actual network server. Whichever result arrives first is used - the other request is canceled (which would cause a log message about a "Broken pipe", as with other connection aborts).
The feature can be disabled in Firefox in about:config by setting network.http.rcwn.enabled to false. For me, I stopped getting log messages after setting this to false. Of course this is not a solution, but the log messages seem to be benign in our case and could be ignored.
For more information about this Firefox feature, see https://slides.com/valentingosu/race-ca ... twork-2017
I believe a future version of Firefox might also mark these "racing" requests in the network panel, which would make identifying problems like this a lot easier.