zentol commented on pull request #11839: URL: https://github.com/apache/flink/pull/11839#issuecomment-619942989
First up, please remember to push updates to a PR as separate commits, so I can see easily which comments have been addressed. > Using the tee will make us have to use the grep to filter out the log4j loggings. Just as the discussion in the ticket. As I said: > As I understand it the only problem we had with the tee approach was that the logging output also ends up in .out . I'm wondering if this is really such a big problem; if we define the .out output as what would be printed to the console, then this is perfectly acceptable behavior. Moving on: > Not all entrypoints need to do redirect the streams. Currently, i think only the docker environment deployment need(e.g. standalone, K8s). This both implies different behaviors depending on how you deploy Flink, and can introduce subtle issues if the supported entry points change. This is exactly the kind of inconsistent behaviors we shouldn't be introducing. > As you have said that the logging framework will cache the OutputStream at creation time, could you share me more insights here? Unfortunately no, but I have been bitten by this in the past. ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: [email protected]
