bryancall commented on issue #11822: URL: https://github.com/apache/trafficserver/issues/11822#issuecomment-4828270361
This was a false-positive crash log: the `traffic_crashlog` helper is parked at startup and armed to be resumed when `traffic_server` exits, so it also woke up on a clean shutdown or restart. It then used a racy `getppid()` check to decide whether a real crash had happened, which meant a normal stop/restart could write an empty `crash-*.log` (note the "No target signal information" and "No target CPU registers" lines, and the target process showing state S/sleeping or Z/zombie, all signs that there was no actual crash). The `traffic_server` process itself was running fine the whole time. I fixed this on master in [PR #13296](https://github.com/apache/trafficserver/pull/13296). The helper no longer guesses from `getppid()`; instead `traffic_server` writes the signal number to it over a socket on a real crash, and on any other exit the socket simply closes so the helper reads end-of-file and exits quietly without writing a log. A regression test was added to cover the clean-shutdown case. The fix is merged with milestone 10.2.0, which has not been released yet, so it is not in 10.0.x or 10.1.x. Closing as resolved on master; please reopen if you still see spurious crash logs on a build that includes this change. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
