During migration from hadoop2 to hadoop3 we noticed a couple of issues with pid files and init scripts. It boils down to the following issue:
The upstream init/systemd scripts expect a pid file name like "yarn-yarn-nodemanager.pid", while our Hadoop 3.1.x installation creates files named "hadoop-yarn-nodemanager.pid". Dealing with the PID directory is straight forward and quite well documented but I cannot find environment variables or run parameters that decide, which file names exactly are used for pid files.
Can somebody point me to the process that does the pid writing? I had the issues with systemd and init.d, so there must be some sourced configuration snippet responsible.
