Viktor's ramblings

In my post How to fix kernel_task CPU usage on Yosemite, I first wrote about how a broken logic board can trigger high CPU usage from kernel_task as well as how to fix it. When El Capitan later were release, the issue remained.

Now, with macOS Sierra out, I upgraded my old MacBook Pro (8,2) as well. As expected, the issue remained the same.

Luckily, the same fix that I wrote about El Capitan still worked.

The only issue I had was that my system didn't boot properly into recovery mode. Instead, I had to utilize a USB stick with the macOS Sierra installer on to be able to turn off System Integration Protection (SIP).

(I try hard to stay away from Windows environments, but sometimes it’s unavoidable.)

Recently I ran across a strange issue with a HP LaserJet 500 colorMFP M570dn. The printer worked great for years, then all of the sudden, it started to act up. Applying the latest firmware didn't solve the issue, so it started probing.

The issue was that the printer randomly went into Offline mode, blocked all users from printing.

After some probing, it turns out that the issue was related to SNMP. In short, Windows tries to determine if the printer is offline using SNMP. This apparently is a common issue.

I've been working on Provisioner for a while now and it's starting to come together. Most recently I rolled out complete API documentation as well as a brand spanking new website (powered by Jekyll and Github Pages).

I'm still trying to think of more use cases, so that's part of what I will be working on next, as well as better documentation at large. I've also started to work on a Python library for easier integration work.

Here are some use cases that I'm thinking of:

Bootstrap remote servers (original concept)

CI/CD assistance for pushing out playbooks to remote servers (store SSH Keys in Provisioner)

Perhaps integrating with Hashicorp Vault?

IFTT or similar integration (?)

I recently had to migrate a number of VMs currently running on an NFS share to an iSCSI target. During my research, I was surprised how little documentation there was around this, so I decided to whip up this quick little piece about how to do it.

Here are the steps:

Create a new iSCSI target (one per VM/image you're migrating). This step varies depending on your setup.

Attach target on server (I found virt-manager suitable for the task).

Find the device from logs (/dev/sdc below). On Ubuntu, you will also need to install the package open-iscsi.

Run qemu-img convert /path/to/your/image.qcow2 -O raw /dev/sdc (for raw images, you could simply use dd but to keep things consistent, I opted for qemu-img).

Update the disk config in the VM definition to point to the iSCSI target.

Boot the VM and archive the old disk image.

Sometime ago, I wrote the blog post How to fix kernel_task CPU usage on Yosemite. This post still receives a great amount of traction, so I wanted to post an update that reflects the covers how to do this on El Capitan.

The process is largely the same, but requires a bit more work due to the changes to the additional security that El Capitan introduced to the file system with System Integration Protection (SIP).

The tl;dr is as follows:

Boot up the system in Recovery Mode (Cmd+R on boot). Start a Terminal window and run csrutil disable.

Reboot the system as normal and follow the same steps as in the original guide.

Reboot the system again into Recover Mode and enable SIP by running csrutil enable.

Reboot the system.

Step 1: Disable System Integration Protection (SIP)

First, shut down your computer. Then power the computer on and boot it into Recovery Mode by holding down Command + R.

Once the computer is done booting, bring up a Terminal window (Utility -> Terminal). With that done, simply run the following command:

$ csrutil disable

That will disable SIP. In order for this to work, you now need to reboot your computer into regular mode (i.e. not anoter Recover Mode boot).

Step 2: Fix the issue

Once your computer is booted, stara Terminal session and run the following commands (for more information, see the original post:

# Find the model $ system_profiler -detailLevel mini | grep "Model Identifier:" Model Identifier: MacBookPro8,2 # Move and backup the file $ cd /System/Library/Extensions/IOPlatformPluginFamily.kext/Contents/PlugIns/ACPI_SMC_PlatformPlugin.kext/Contents/Resources $ sudo mv MacBookPro8_2.plist MacBookPro8_2.bak

With that done, it's time to go back into Recover Mode again, so shut down your computer.

Step 3: Re-enable SIP

Boot the computer in Recover Mode again by pressing Command + R on boot. Again, open a Terminal window, but this time, run the following command:

$ csrutil enable

Now reboot your computer and you should be all set.

For some time, Ansible has been my configuration management of choice and we use it for both Screenly and YippieMove. Since both of these services are running on Google Compute Engine, we're using Ansible's dynamic inventory for GCE.

(Behind the scenes, this dynamic inventory is using Apache Libcloud, which is a great Python library for interacting with various providers.)

When I first followed Ansible's Google Cloud Platform Guide I did run into a fair bit of trouble with authentication.

As it turns out, Libcloud is very picky when it comes to the environment variables that you need to set. To solve this, I whipped together a little script that I call on to set the appropriate variable environments for each project I'm working on. This saved me a lot of headache.

The script looks as follows:

#!/bin/bash export GCE_PROJECT=your-project export GCE_PEM_FILE_PATH=~/.gce/$GCE_PROJECT\.json export GCE_EMAIL=$(grep client_email $GCE_PEM_FILE_PATH | sed -e 's/ "client_email": "//g' -e 's/",//g') gcloud config set project $GCE_PROJECT

Just change GCE_PROJECT to match your setup, and then run:

$ source /path/to/script.sh

You can now run Ansible with the GCE inventory file.

As an added bonus, this also configures gcloud to the same project.

The built-in Internet Sharing in OS X is very handy. When I'm on the road, I frequently use this to share my laptops WiFi connection with other devices over a wired connection (such as a Raspberry Pi, when I work on Screenly).

If you're connecting some kind of headless device, you will likely want to connect to this device over SSH or similar. The only problem is that you don't know the IP address of said device (it's headless, remember).

Luckly, all the tools you need are already available on OS X. All you need to do is to fire up the Terminal:

I'm a huge fan of autojump. It allows me to quickly navigate my filesystem in ways without having to type out every folder.

There is however one task that I frequently that I wanted to make more efficient: create a new folder and then jump into said folder.

Normally, this would simply be:

$ mkdir foo $ cd foo

This feels somewhat inefficient, so I wrote a little tool to help with this called ccd:

$ ccd foo

The tool is very simple, but saves me a number of keystrokes every day. There are two pieces to the tool: one bash script and one entry in ~/.profile.

~/bin/ccd.sh

#!/bin/bash ARG="$1" if [ ! -d "$ARG" ]; then echo "Creating $ARG." mkdir -p "$(pwd)/$ARG" else echo "$ARG already exists." fi cd "$(pwd)/$1"

Once you have installed the script, just set the right permission with chmod +x ~/bin/ccd.sh

~/.profile

Lastly, you will need to add the following entry to your ~/.profile file:

alias ccd="source ~/bin/ccd.sh"

After moving to London, I decided to go with NatWest as my bank. While it was a good experience at large, their online banking leavs a lot to be desired. Not only do they lack things like Two-Factor Authenciation (2FA), but they also have this really frustrating login system.

Well, today I had enough and whipped up a CLI tool that allows you to generate the data that they ask for by reading it in from 1Password (using 1pass).

Dear Digital Security Controls (DSC),

I recently purchased your T-link extension (IP module) for two alarm systems that I wanted to be able to remotely manage. After a few hours of troubleshooting (read: having to install Windows XP), I am now able to connect to these alarm systems over a VPN connection, but that's about it.

Please take a look at this screenshot:

If it isn't obvious what is wrong with this, let me spell this out for you.

Your software only runs on Windows XP. Nope, not even Windows 7. Forget about a modern web interface like most people have come to expect in this day and age from an 'IP module'.

Your client software requires SQL Server Express (ehh WTF?!).

Throughout the entire app, you use the industry standard icon that means 'refresh' for 'Restore all options as to original'. I just almost hit this icon when I was about to reload the values.

In a client software download means just that, down load (i.e. fetching to your local computer) and upload means to upload from the computer to whatever device/service you're intereacting with. In your software, you've apparently decided to call them the other way around (which is very confusing).

The user interface is confusing to say the least. I've worked with a few different alarm systems in the past. They've all been ranging from crappy to less crappy, but this one takes the price. After spending a good 15 minutes, I still have no idea how to perform even basic tasks (like user management, set PIN codes, look at events etc).

I could go on and on about this, but really, I don't have enough energy. Anyone with any UI/UX experience can show you. I'm just severely disappointed that some has the stomach to sell something this poorly designed in this day and age. It should barely be classified as a Minimum Viable Product (MVP).

If you want to be in the software space, please hire people who are actually qualified for the task.