https://bz.apache.org/bugzilla/show_bug.cgi?id=57832
Yann Ylavic <[email protected]> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |[email protected] --- Comment #3 from Yann Ylavic <[email protected]> --- Created attachment 32698 --> https://bz.apache.org/bugzilla/attachment.cgi?id=32698&action=edit Proposed patch on top of r1656259 An effort has been made in r1656259 (trunk only) to reduce the time between the backend connection is checked for availability/readability and if that's OK then reused for the next request. Thus I think the attached patch (on top) should be enough to address this potential issue. The check is done in ap_proxy_connect_backend() because all proxy modules (reusing connections) currently handle/assume a transactional protocol (one request leading to one response for HTTP, AJP, ...). Since r1656259, ap_proxy_connect_backend() is called just before sending the (next) request, so it looks like the right place to fail if some data are readable. Can you give that a try (applying both r1656259 and the attached patch to 2.4.x seems to work too)? -- You are receiving this mail because: You are the assignee for the bug. --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
