https://bz.apache.org/bugzilla/show_bug.cgi?id=64400

--- Comment #10 from Felix Schumacher <[email protected]> ---
Created attachment 37216
  --> https://bz.apache.org/bugzilla/attachment.cgi?id=37216&action=edit
Possible race condition when placing samplers

Alex, I think the problems arise from a race condition inside the placement
code for the recorded samples. I tried to extract the logic to be less prone to
this race condition. Could you try the patch with your setup?

There is at least one place, where we still are not 100 % thread safe. That is
in ProxyControl#deliverSample when we compute the target node 'myTarget'. That
happens outside the Swing Thread and thus could result in the wrong target,
too.

To get a wrong result with the current code base, you have to be fast (changing
the name of the transaction controller) and have a lot of concurrent (or near
concurrent) requests. I used a shell command 'for i in $(seq 200); do curl
http://targethost & done; wait'

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to