SolidWallOfCode opened a new pull request, #9114: URL: https://github.com/apache/trafficserver/pull/9114
This is based on a problem I had at LinkedIn involving latency in requests. The event loop data looked as if the event loops were very long (billions of nanoseconds in some cases) but it wasn't possible to tell if this was the exception or the rule (i.e. is there a plugin that now and then blocks on an `ET_NET` thread, or are the plugins in general slowing things up?). This adds histogram data for event loop timing so it's possible to see 1. How relatively common long loops. 2. Whether there are many long loops, instead of tracking only the very longest. 3. Whether there are clusters at certain timings. This should be backwards compatible with 9.x. As part of this work, the `Histogram` class is added. It is designed to log samples performantly and provides a decay mechanism to keep the data relevant. These call all be adjusted by tweaking various compile time constants. -- 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]
