Users of the Nagios NRPE plugin will soon be finding that even when properly configured, it consistently fails to start on boot – even though it can be manually started later, and everything looks good.

The problem is that Nagios need to update their systemd startup unit files with the proper dependencies – to ensure the NRPE server isn’t started before the network is available.

Currently, the best solution is a systemd drop-in:

Create directory: /etc/systemd/system/nrpe.service.d
Create a new drop-in file, at /etc/systemd/system/nrpe.service.d/nrpe.conf

[Unit]
Requires=network-online.target
After=network-online.target


systemctl daemon-reload
systemctl restart nrpe

Now check the NRPE server status and you’ll see the service should be starting successfully, with the drop-in showing:


[root@voltaire 01-02 22:32:50 ~]# systemctl status nrpe
- nrpe.service - Nagios Remote Program Executor
Loaded: loaded (/usr/lib/systemd/system/nrpe.service; enabled; vendor preset: disabled)
Drop-In: /etc/systemd/system/nrpe.service.d
|-nrpe.conf
Active: active (running) since Tue 2018-01-02 22:32:43 ACDT; 13s ago
Docs: http://www.nagios.org/documentation
Main PID: 1392 (nrpe)
CGroup: /system.slice/nrpe.service
|-1392 /usr/sbin/nrpe -c /etc/nagios/nrpe.cfg -d
.
Jan 02 22:32:43 voltaire.ovirt.vorpal systemd[1]: Starting Nagios Remote Program Executor...
Jan 02 22:32:43 voltaire.ovirt.vorpal systemd[1]: Started Nagios Remote Program Executor.


You can see the drop-in has successfully loaded in the 3rd/4th lines of output above.

This should ensure the NRPE service starts successfully every time you reboot.

Over the Xmas break I was also busy beavering away finishing off my Jenkins CI pipeline to automate building VTK.

VTK (Visualisation ToolKit) is an excellent library of code for processing and visualisaton of multi-dimensional datasets. It’s written in C++ for high performance, and comes with bindings for python, java, and tcl/tk. For more info, see the VTK website

I’ve done a bit more of a write-up here

The jenkins-vtk-pipeline project is here, on GitHub

Have been beavering away over the Xmas break, getting the beginnings of an SOE up and running using Ansible for the initial setup.

The plan is probably to switch to Puppet for most of the subsequent setup, but Ansible is a good tool especially for the early stages.

Currently targeting Red Hat Enterprise Linux (RHEL)/CentOS/Fedora & Ubuntu.

Have done a project page which I’ll add more to soon, here.

Project is here, on GitHub.

Recently, after upgrading my main system to Fedora 27, I found my vncservers were intermittently failing to start on boot.

Status messages indicated the pam_nologin module was involved, and servers were failing to start because /var/run/nologin had not been removed yet.

For anyone else who hits this issue, all it means is things are starting up faster on F27, and systemd is trying to start your vncservers before the system is ready for user logins (derp).

The fix is easy, just add systemd-user-sessions.service to the “After” line in the systemd vncserver template file, and in any vncserver units you’ve already created from your template file.

So your template file should look something like this:


[Unit]
Description=Remote desktop service (VNC)
After=syslog.target network.target systemd-user-sessions.service

#Notes:
#-fg is necessary so vncserver stays in foreground, otherwise, the systemd user session shuts down as
#soon as vncserver forks off its daemon process and the parent exits.