this is probably *related* to lp bug #1568650 - but not exactly the same.
The same repo script [1] with a significantly slow enough client (> 60 seconds by default) will trigger the ChunkWriteTimeout. The script as written just simulates a really *slow* client; not necessarily bad behaved and trying to wreck things - but with high enough sleep you could simulate a Slowloris/DOS if that's what you're going for.
In my most recent test the proxy seemed to square it's backend connections, but the socket to the client remained established.
this is probably *related* to lp bug #1568650 - but not exactly the same.
The same repo script [1] with a significantly slow enough client (> 60 seconds by default) will trigger the ChunkWriteTimeout. The script as written just simulates a really *slow* client; not necessarily bad behaved and trying to wreck things - but with high enough sleep you could simulate a Slowloris/DOS if that's what you're going for.
In my most recent test the proxy seemed to square it's backend connections, but the socket to the client remained established.
1. https:/ /gist.github. com/clayg/ 2458a4a7e1451c7 5fbc5a63fcae116 35