Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
[jetty-users] Bug between 9.2.3 and 9.2.5


I'm chasing a bit of a strange bug. I'm wondering if anyone can suggest how I should go about generating a reproducible test case.

I've got one particular response which is almost entirely sent, then there's a delay of approx 30 seconds, then the final bit arrives and the page rendering is completed. It happens the same in Firefox, Chrome and Safari.

It's from an embedded Jetty 9.2.5 returning a web page which is 438368 bytes uncompressed, 71418 bytes compressed, with about 262 bytes of headers, over an SSL connection.

The gzipping is done by our app, not jetty. It's sent by calling servletResponse.getOutputStream() then write() on the stream via a GZIPOutputStream with an 8k buffer.

I'm almost certain the write() returns quickly, and as far as the app is concerned, it finished handling the request in a few milliseconds.

It's pretty much 100% repeatable.

Reverting to 9.2.3 solves the problem.

I'm running on a Solaris derivative:

   java version "1.7.0_71"
   Java(TM) SE Runtime Environment (build 1.7.0_71-b14)
   Java HotSpot(TM) Server VM (build 24.71-b01, mixed mode)

I can't seem to reproduce this outside our production environment with data I can't share. I've tried generating all sorts of responses of about this size, and even using the text of the offending response, all to no avail.

Any suggestions welcomed!



Back to the top