Gitlab Setup and Timeouts

So I wanted a place to store all my private Git repositories, without going down the paid route that is Github. Callum had just set up Gitlab at work, so I thought it might be wise to just use the same toolset.

What with it being installed on a VM at home, it didn’t quite have the same power, but that shouldn’t be too much of an issue. However, after the install and configuration check, I was plagued with Nginx gateway timeouts, blank pages, and no immediately apparent error messages other than the Nginx ones – until I looked at unicorn.stderr.log. The Gitlab process was timing out – yet the configuration check said all was good. Database connectivity was there, as was networking, etc. What gives?

This post however, suggested another problem – Unicorn timeouts. The process isn’t being given long enough to initialise assets, and is timing out. Simply increasing the ‘timeout’ parameter in config/unicorn.rb, restart Gitlab, and give it a moment.

Automated Kitchen Lighting

Recently I decided to make some changes to the kitchen. It started off with a strip of warm white LEDs, and I thought “Wouldn’t it be cool if…”. I’ve now lost count of the number of guests asking how it works 🙂

There’s a variety of components at play here. Firstly, there’s a Raspberry Pi, followed up by a simple IRF530 MOSFET, a PNP transistor to pull it low, a PIR sensor, and a strip of LEDs.

Kitchen lighting


It’s a a PIR sensor from Seeedstudio that I bought on a whim a while back. It handily runs off of 5v, and returns a 3.3v TTL compatible signal upon detection of movement.

On the Raspberry Pi, I have my own Python app, mqtt-gpio-trigger, running. It continuously monitors pins that are defined in the config file, and in the event of a change of state of any of them it will publish a message to a predefined topic, with the pin number as the final component of the topic, and the message contains the state of the PIN. Simply subscribe to that topic to watch whatever pin on the Raspberry Pi.


A strip of warm white LED lights are attached to pin 18 of the Raspberry Pi, via the IRF530 MOSFET. There’s a bog standard NPN transistor to help saturate the gate of the MOSFET, and it’s pin 18 as that’s the only pin on the Raspberry Pi that does hardware PWM. Using any of the other pins subjects you to vagaries of the CPU.

Controlling pin 18 is mqtt-pwm-lights, another one of my scripts. It simply sits and runs, and subscribes to yet another MQTT topic. When it receives a message with a value between 0 and 512, it fades the lights to that value. Due to the NPN transistor it’s inverted, so 512 is off, and 0 is on. A minor annoyance, but the home automation controller handles that.

In the middle

Producing the show is a small script that isn’t public. All it does is listen for MQTT events and reply with canned MQTT messages in return. It’s a stopgap measure until I find a better home automation controller. The main role of the script at the moment is to listen to events from the PIR, start a timer when triggered, and publish a lighting level request. When the timer expires, the lighting level returns to normal (ie, off)

So that’s it. Ambient automated lighting for the kitchen, and worktop lighting, all rolled into one.

Thinkpad X201S Fan Control

I noticed the other day (it was kind of hard to not notice, to be honest) that my laptop shut down abruptly when it overheated. This shouldn’t be the case, as the thermal controls on the mainboard should have spun up the fans.

After some poking around, I found that with the default ‘auto’ fan control, the top speed of ‘level 7’ only spins the fans around 4000RPM, even though when the feedback loop is disengaged (echo level disengaged > /proc/acpi/ibm/fan” they can spin up to 7000RPM, and bring the temperature down from the critical 100C mark.

This bug report suggests that something is completely amiss with the kernels thinkpad_acpi module, and nobody really knows why. The end result is that fans spin too slow on the Thinkpad range of laptops. This can, however, be resolved with a couple of edits and an additional package. The following instructions were tested on Ubuntu 12.10, most likely work exactly the same on all Debian versions and derivatives, and can be adapted for use on other Linux distros as well.

Firstly, enabled userland control of the Thinkpad fans with the following command –

echo options thinkpad_acpi fan_control=1 >> /etc/modprobe.d/thinkpad.conf
rmmod thinkpad_acpi
modprobe thinkpad_acpi

This will enable the new settings for your current session, and all future sessions too.

You should now be able to set the fan speed manually with commands such as –

echo level 0 > /proc/acpi/ibm/fan
echo level 7 > /proc/acpi/ibm/fan
echo level disengaged > /proc/acpi/ibm/fan
echo level auto > /proc/acpi/ibm/fan

The key bit here is to remember that ‘disengaged’ equates to ‘127’ on the grand scale of speeds.

Next step is to install the thinkfan package, and enable automatic startup –

apt-get -y install thinkfan
sed -i s/START=no/START=yes/ /etc/default/thinkfan

Now, as above, the default fan speeds are too slow. So even if thinkfan commands a level 7 fan speed, it won’t be fast enough to keep your hardware cool. To make it run up to full speed, you have to disengage the feedback loop and let it run in failsafe mode. Don’t worry though, thinkfan will bring it back into ‘auto’ mode once temperatures are back to normal, so you can have normal battery life and normal noise levels once your compute process is complete.

Edit /etc/thinkfan.conf, and at the end adjust the endpoint for level 7, and insert a new line for fan speed 127. My thinkfan.conf is as below –

(0, 0, 55)
(1, 48, 60)
(2, 50, 61)
(3, 52, 63)
(4, 56, 65)
(5, 59, 66)
(7, 63, 66)
(127, 66, 32767)

Save thinkfan.conf, and start thinkfan with this

/etc/init.d/thinkfan start
Once 66C is passed, it will enter disengaged mode, and run at full speed until it reaches 66C or below again.

My aging quad core i7 2.13GHz laptop will happily run BOINC now, at 97C, so I say it’s working 🙂

OWFS Server on The Raspberry Pi

Quick and simple this one is. I have a DS9490R (internally it’s a DS1490F) one-wire bus master attached to my Raspberry Pi. 4 DS18B20 temperature sensors are hanging off the bus, and I’d like to access them from the home automation server.

On the Raspberry Pi, or the server/device that’s hosting the 1-Wire bus:

apt-get install owserver ow-shell
modprobe ds1wm
echo “ds1wm” >> /etc/modules

Edit /etc/owfs.conf such that the following parameters are entered. I’d advise commenting out the rest (at least, the rest of the server: directives)

! server: server = localhost:4304
server: usb = all
server: w1
server: port =

Restart the 1-Wire server

/etc/init.d/owserver restart

All being well, the following command should output a listing of all the devices on the bus.

owdir -s localhost:4304 /

On the remote server, simply use the following command to read the data.

owread -s /28.DDBF1D030000/temperature

MQTT GPIO Monitoring

My little Raspberry Pi seems to be growing arms and legs.

The other day I hooked up a simple PIR to it. I can’t remember where I got it, but it runs off 5v, consumes a low enough amount of milliAmps to be directly connected to the Raspberry Pi 5v supply, and outputs a TTL compatible 3.3v signal. Bonus!

I wanted to be able to signal to my MQTT broker when motion was detected, so it was time to start monitoring the GPIO pins. I wrote up this short app using the framework I’d made for MQTT-Republisher earlier. MQTT-GPIO-Trigger will accept a list of GPIO pins that you wish to watch, and will cycle through them all and fire off MQTT messages on any state change.

It uses the standard Linux sysfs interface to read the pin states, and if it detects the WiringPi library on the system it will use Gordons gpio command to export the pins to the system.