pcs trying to use systemctl even though sysvinit is the init system

[jpraczyk]

System: Debian 8 Jessie

In Debian's case it is perfectly fine to have multiple init systemd installed at the same time.
The is_systemctl check performed in lib/external.py assumes that if there are systemd's binaries installed in the system, then it must be running systemd as init system.

Steps to reproduce:

1. Install sysvinit-core
2. Reboot the system
3. Try starting the cluster
4. Receive the following error when running pcs cluster start:

# pcs cluster start
Starting Cluster...
Failed to get D-Bus connection: Unknown error -1
Error: unable to start corosync
# /etc/init.d/corosync status
corosync is running.

To workaround this issue, one could simply remove systemd from the system, but it's inconvenient.

# apt remove systemd -y
# pcs cluster start
Starting Cluster...

Proposed solutution:
Check if systemd is used as the init system. Sysvinit will always have pid = 1, whereas in systemd's case it appears to be somewhat random:

$ pidof systemd
6906

# pidof systemd
# pidof init
1

[tomjelinek]

Using pidof is not reliable. On RHEL7 pidof systemd returns 1.

Can you check if /run/systemd/system or /var/run/systemd/system exist when your init system is systemd and not exist when your init system is SysVinit?

[vvidic]

The standard check for a running systemd in the packaging scripts is:

if [ -d /run/systemd/system ]; then ...

so the same directory check should work in python too.

[tomjelinek]

Fixed in fbfe34f and 604692f