Hide Forgot
Description of problem: When clurgmgrd launches a service it forks and calls execle. this creates a new process in the same process group. If the new process issues a signal to the process group such as kill(0, SIGTERM) then the entire process group responds including clurgmgrd. Version-Release number of selected component (if applicable): rgmanager-2.0.52-9 How reproducible: Consistently Steps to Reproduce: 1. Create a service that uses the kill(0, SIGTERM) and start it. 2. 3. Actual results: clurgmgrd receives SIGTERM Expected results: clurgmgrd should not shut down Additional info:
Created attachment 492988 [details] Test case This unit test can be attached as a script to a service. Place this service script on all nodes and simply create a service which looks like the following: <service name="test"> <script name="pgtest" file="/root/pgtest.sh" /> </service> Ensure you have gcc installed on all hosts, as the script compiles a quick bit of C code to run. If run on a version of rgmanager exhibiting this problem (child processes in the same process group), the script will refuse to start and log an error in the system logs. On a system without the process group problem, the script will start without issue. Running this test script from rgmanager is likely to require SELinux being disabled or at least placed in permissive mode.
Created attachment 492989 [details] Fix
Development Management has reviewed and declined this request. You may appeal this decision by reopening this request.