Github user shinrich commented on the pull request:
https://github.com/apache/trafficserver/pull/145#issuecomment-63975572
We ended up leaving in the kill_tunnel in HttpSM::kill_this(). TS-1527 is
fixed by this addition as well as the issue I was seeing here. The possibility
of double do_io_closes() is disturbing, but that seems to happen elsewhere as
well. If it is in the same invocation stream it is safe (just resetting the
close flag). Should be investigated in more depth though. Based on what we
were seeing in production and the little bit of gdb investigation that Alan
did, the keep_alive connections are not affected, but again, we should
investigate in more detail and perhaps add a test case to the integration test
set to verify that shutting down the SM does not destroy the keep alive
connections.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
