[
https://issues.apache.org/jira/browse/TS-4503?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15312994#comment-15312994
]
ASF GitHub Bot commented on TS-4503:
------------------------------------
Github user shinrich commented on the issue:
https://github.com/apache/trafficserver/pull/694
In our most recent fire drill, the core was very useful. The particular
problem only happens under very heavy load. By poking around the core, we got
some idea of the underlying cause and were able to add additional checks to
attempt to catch the problem next time around.
The problem was that the reported stack was showing secondary errors from
the memory cleanup. So we were looking in the wrong area for several days
until we got access to the machine and core. Then staring at the core for a
while until it occurred to me to do "info thread" and find the real problem.
> MachineFatal should shutdown without cleanup
> --------------------------------------------
>
> Key: TS-4503
> URL: https://issues.apache.org/jira/browse/TS-4503
> Project: Traffic Server
> Issue Type: Bug
> Reporter: Susan Hinrichs
> Assignee: Susan Hinrichs
> Fix For: 7.0.0
>
>
> When MachineFatalClass::raise() is called, it prints a message to the log and
> then calls exit. exit causes memory cleanup to be called. But if we are in
> such bad state that MachineFatal is called, it is quite likely that memory is
> messed up.
> We saw a crash where MachineFatal is called in thread 84. This stack has the
> real error. But the stack that got reported was on thread 1 in class
> destructor logic. It would have been better if ATS failed immediately and
> the stack on thread 84 was reported.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
