Setting up a PR to revert this commit. After we ran this change in production for a while, we noticed higher TCP_MISS 500's and increased average response time. Looking more closely into the cores, the locks were not being held, but the bigger issue was that the continuations themselves were stale. Other commits (that have been committed or set up as open source PRs) seems to have addressed this issue. We haven't seen this stack trace in later builds.
[ Full content available at: https://github.com/apache/trafficserver/pull/4104 ] This message was relayed via gitbox.apache.org for [email protected]
