Foreman » Packaging

Cloned from https://bugzilla.redhat.com/show_bug.cgi?id=1628505

Description of problem:
We used systemctl kill ... to deliver SIGUSR1 to smart_proxy_dynflow_core to make it reopen its logs. While systemctl kill delivered the signal reliably, it delivered the signal to every process in the service's control group. The issue was when the service forked more processes (when running ansible-playbook). pkill signals only the processes with matching name.

Ansible-playbook reacts to SIGUSR{1,2} by printing "User defined signal
1" and exiting, which is undesired.

Version-Release number of selected component (if applicable):
6.4

How reproducible:
Always

Steps to Reproduce:
1. Run sleep 1000 using the ansible command template
2. When the job runs, run logrotate -f /etc/logrotate.conf -v
3. (optional) Watch the logrotate output as it may not rotate the logs if they were rotated recently

Actual results:
Logs are rotated, ansible processes die. Possible IOError shown in the jobs output

Expected results:
Logs are rotated, running ansible processes are left intact.

Additional info: