In today’s installment:

• dmesg from before the machine crashed
• Kill process ID from pidfile
• Let systemd retry a task
• Set MTU on OpenVPN connections in Networkmanager

dmesg from before the machine crashed

Super helpful and complete after most kernel panics and Magic SysRq resets:

journalctl --dmesg --boot -1

Kill process ID from pidfile

pkill --pidfile foo.pid

I use this in this wacky systemd user unit:

[Unit]
Description=nginx restreamer

[Service]
ExecStart=/usr/sbin/nginx \
        -c %h/.local/nginx/nginx.conf \
        -g "pid %h/.local/nginx/nginx.pid;"
ExecReload=pkill -HUP --pidfile %h/.local/nginx/nginx.pid

[Install]
WantedBy=default.target

Let systemd retry a task

Not the first appearance of systemd-run in this category, and probably not the last. Here’s how to hand a task to systemd-run in order to be retried a number of times.

My bro math for interval calculation goes like this:

• RestartSec = Time to wait between end of a failed attempt and start of the next attempt
• RuntimeMaxSec = Allowed time for each attempt
• StartLimitBurst = Number of attempts to make
• StartLimitIntervalSec = (RestartSec + RuntimeMaxSec) x StartLimitBurst x 10

So for a service that shall retry every 5 minutes (RestartSec), 12 times (StartLimitBurst), and shall be timed out after 30 seconds (RuntimeMaxSec):

systemd-run \
    --user \
    --property=Restart=on-failure\
    --property=RestartSec=300 \
    --property=RuntimeMaxSec=30
    --property=StartLimitBurst=12 \
    --property=StartLimitIntervalSec=39600 \
    sh -c 'test -e /tmp/1 && touch /tmp/1'

Set MTU on OpenVPN connections in Networkmanager

Is this really still required?

sudo install -m 0755 /dev/stdin /etc/NetworkManager/dispatcher.d/vpn-up << "Here"
#!/bin/sh

if [ "$2" = "vpn-up" ]; then
        ip link set dev "$1" mtu 1392
fi
Here