RHEL7/Oracle HA: systemd spends its time sending "kill0" to processes and reading "/proc/PID/stat" files, causing high CPU consumption

Issue

• On RHEL7 system running oracle-ohasd.service unit, high CPU consumption is seen for systemd process

  USER       PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
  root         1 36.1  0.0 198320 11308 ?        Rs    2021 247455:13 /usr/lib/systemd/systemd --switched-root --system --deserialize 22
  

• Stracing systemd (PID 1) for 30s shows a very high number of kill(<PID>, SIG_0) operations and opening/reading of /proc/<PID>/stat nodes

  # timeout 30s strace -ttTvyy -s 128 -o /tmp/systemd.strace -p 1
  
  # egrep -c "kill\([0-9]+, SIG_0\)" /tmp/systemd.strace
  8293
  
  # egrep -c "open\(\"/proc/[0-9]+/stat\"," /tmp/systemd.strace
  8308