Finally figured out what was going on and why the ora.service script was not working. It was not working because ASM processes were being killed immediately at start by systemd. The issue was caused by configuration option “RemoveIPC” in logind daemon‘s configuration file /etc/systemd/logind.conf since the option was enabled by default:
#RemoveIPC=no
What a disaster! Note that the line is commented (“#”). That means RemoveIPC=yes, because “yes” is default. The line must be uncommented
RemoveIPC=no
and after reloading and restarting logind daemon
systemctl daemon-reload
systemctl restart systemd-logind.service
the issue was solved.
I don’t have a clue about the real purpose and real motive of introducing this option in systemd, but it seems that all the processes owned by users whose User ID > 50000 are silently killed without notification. I was just little confused why that was not happening on my other machine with Fedora 25 Server and Oracle 12c without Grid where database was always being started successfully with similar ora.service script. Just checked out and found that the RemoveIPC default value is changed to “no” in Fedora 25, given that the commented line now reads #RemoveIPC=yes. Really messy. The systemd version in Fedora 25 is 231-12, while Oracle Linux 7.3 has 219-30.
My claim that it was the systemd issue, not the Oracle Restart nor Database Installation issue, is hereby proven as the fact.
To Avi Miller. Given that I was right and you were wrong, hope you will be honest enough to move my original thread from Database Installation forum back to Oracle Linux forum. It was deleted there, but they still have it in their Recycle Bin. Thank you.