major.io words of wisdom from a systems engineer

Automatic floating windows in i3

floating leaf

The i3 window manager is a fast window manager that helps you keep all of your applications in the right place. It automatically tiles windows and can manage those tiles across multiple virtual desktops.

However, there are certain applications that I really prefer in a floating window. Floating windows do not get tiled and they can easily be dragged around with your mouse. They’re the type of windows you expect to see on other non-tiling desktops such as GNOME or KDE.

Convert a window to floating temporarily

If you have an existing window that you prefer to float, select that window and press Mod + Shift + Space bar. The window will pop up in front of the tiled windows and you can easily move it with your mouse.

Depending on your configuration, you may be able to resize it by grabbing a corner of the window with your mouse. You can also assign a key combination for resizing in your i3 configuration file (usually ~/.config/i3/config):

# resize window (you can also use the mouse for that)
mode "resize" {
        bindsym Left resize shrink width 10 px or 10 ppt
        bindsym Down resize grow height 10 px or 10 ppt
        bindsym Up resize shrink height 10 px or 10 ppt
        bindsym Right resize grow width 10 px or 10 ppt
        bindsym Return mode "default"
        bindsym Escape mode "default"
        bindsym $mod+r mode "default"
}
bindsym $mod+r mode "resize"

With this configuration, simply press Mod + r and use the arrow keys to grow or shrink the window’s borders.

Always float certain windows

For those windows that you always want to be floating no matter what, i3 has a solution for that, too. Just tell i3 how to identify your windows and ensure floating enable appears in the i3 config:

for_window [window_role="About"] floating enable
for_window [class="vlc"] floating enable
for_window [title="Authy"] floating enable

In the example above, I have a few windows always set to be floating:

  • [window_role="About"] - Any of the “About” windows in various applications that are normally opened by Help -> About.
  • [class="vlc"] - The VLC media player can be a good one to float if you need to stuff it away in a corner.
  • [title="Authy"] - Authy’s chrome extension looks downright silly as a tiled window.

Any time these windows are spawned, they will automatically appear as floating windows. You can always switch them back to tiled manually by pressing Mod + Shift + Space bar.

Identifying windows

Identifying windows in the way that i3 cares about can be challenging. Knowing when to use window_role or class for a window isn’t very intuitive. Fortunately, there’s a great script from an archived i3 faq thread that makes this easy:

Download this script to your system, make it executable (chmod +x i3-get-window-criteria), and run it. As soon as you do that, a plus (+) icon will replace your normal mouse cursor. Click on the window you care about and look for the output in your terminal where you ran the i3-get-window-criteria script.

On my system, clicking on a terminator terminal window gives me:

[class="Terminator" id=37748743 instance="terminator" title="[email protected]:~"]

If I wanted to float all terminator windows, I could add this to my i3 configuration file:

for_window [class="Terminator"] floating enable

Float in a specific workspace

Do you need a window to always float on a specific workspace? i3 can do that, too!

Let’s go back to the example with VLC. Let’s consider that we have a really nice 4K display where we always want to watch movies and that’s where workspace 2 lives. We can tell i3 to always float the VLC window on workspace 2 with this configuration:

set $ws1 "1: main"
set $ws2 "2: 4kdisplay"
for_window [class="vlc"] floating enable
for_window [class="vlc"] move to workspace $ws2

Restart i3 to pick up the new changes (usually Mod + Shift + R) and start VLC. It should appear on workspace 2 as a floating window!

Photo source

DevConf.CZ 2019 Recap

DevConf.CZ 2019 wrapped up last weekend and it was a great event packed with lots of knowledgeable speakers, an engaging hallway track, and delicious food. This was my first trip to any DevConf and it was my second trip to Brno.

Lots of snow showed up on the second day and more snow arrived later in the week!

devconf-snow-small

First talk of 2019

I co-presented a talk with one of my teammates, Nikolai, about some of the fun work we’ve been doing at Red Hat to improve the quality of the Linux kernel in an automated way. The room was full and we had lots of good questions at the end of the talk. We also received some feedback that we could take back to the team to change how we approached certain parts of the kernel testing.

devconf-major-nikolai

Our project, called Continuous Kernel Integration (CKI), has a goal of reducing the amount of bugs that are merged into the Linux kernel. This requires lots of infrastructure, automation, and testing capabilities. We shared information about our setup, the problems we’ve found, and where we want to go in the future.

Feel free to view our slides and watch the video (which should be up soon.)

Great talks from DevConf

My favorite talk of the conference was Laura Abbott’s “Monsters, Ghosts, and Bugs.”

devconf-laura-abbott

It’s the most informative, concise, and sane review of how all the Linux kernels on the planet fit together. From the insanity of linux-next to the wild world of being a Linux distribution kernel maintainer, she helped us all understand the process of how kernels are maintained. She also took time to help the audience understand which kernels are most important to them and how they can make the right decisions about the kernel that will suit their needs. There are plenty of good points in my Twitter thread about her talk.

Dan Walsh gave a detailed overview of how to use Podman instead of Docker. He talked about the project’s origins and some of the incorrect assumptions that many people have (that running containers means only running Docker). Running containers without root has plenty of benefits. In addition, a significant amount of work has been done to speed up container pulls and pushes in Podman. I took some notes on Dan’s talk in a thread on Twitter.

The firewalld package has gained some new features recently and it’s poised to fully take advantage of nftables in Fedora 31! Using nftables means that firewall updates are done faster with fewer hiccups in busy environments (think OpenStack and Kubernetes). In addition, nftables can apply rules to IPv4 and IPv6 simultaneously, depeending on your preferences. My firewalld Twitter thread has more details from the talk.

The cgroups v2 subsystem was a popular topic in a few of the talks I visited, including the lightning talks. There are plenty of issues to get it working with Kubernetes and container management systems. It’s also missing the freezer capability from the original cgroups implementation. Without that, pausing a container, or using technology like CRIU, simply won’t work. Nobody could name a Linux distribution that has cgroups v2 enabled at the moment, and that’s not helping the effort move forward. Look for more news on this soon.

devconf-openshift-pun

OpenShift is quickly moving towards offering multiple architectures as a first class product feature. That would incluve aarch64, ppc64le, and s390x in addition to the existing x86_64 support. Andy McCrae and Jeff Young had a talk detailing many of the challenges along with lots of punny references to various “arches”. I made a Twitter thread of the main points from the OpenShift talk.

Some of the other news included:

  • real-time linux patches are likely going to be merged into mainline. (only 15 years in the making!)
  • Fedora, CentOS, RHEL and EPEL communities are eager to bring more of their processes together and make it easier for contributors to join in.
  • Linux 5.0 is no more exciting than 4.20. It would have been 4.21 if Linus had an extra finger or toe.

DevConf.US Boston 2019

The next DevConf.US is in Boston, USA this summer. I hope to see you there!

Using the pressure stall information interface in kernel 4.20

pressure_cooker

Fedora 29 now has kernel 4.20 available and it has lots of new features. One of the more interesting and easy to use features is the pressure stall information interface.

Load average

We’re all familiar with the load average measurement on Linux machines, even if the numbers do seem a bit cryptic:

$ w
 10:55:46 up 11 min,  1 user,  load average: 0.42, 0.39, 0.26

The numbers denote how many processes were active over the last one, five and 15 minutes. In my case, I have a system with four cores. My numbers above show that less than one process was active in the last set of intervals. That means that my system isn’t doing very much and processes are not waiting in the queue.

However, if I begin compiling a kernel with eight threads (double my core count), the numbers change dramatically:

$ w
 11:00:28 up 16 min,  1 user,  load average: 4.15, 1.89, 0.86

The one minute load average is now over four, which means some processes are waiting to be served on the system. This makes sense because I am using eight threads to compile a kernel on a system with four cores.

More detail

We assume that the CPU is the limiting factor in the system since we know that compiling a kernel takes lots of CPU time. We can verify (and quantify) that with the pressure stall information available in 4.20.

We start by taking a look in /proc/pressure:

$ head /proc/pressure/*
==> /proc/pressure/cpu <==
some avg10=71.37 avg60=57.25 avg300=23.83 total=100354487

==> /proc/pressure/io <==
some avg10=0.17 avg60=0.13 avg300=0.24 total=8101378
full avg10=0.00 avg60=0.01 avg300=0.16 total=5866706

==> /proc/pressure/memory <==
some avg10=0.00 avg60=0.00 avg300=0.00 total=0
full avg10=0.00 avg60=0.00 avg300=0.00 total=0

But what do these numbers mean? The shortest explanation is in the patch itself:

PSI aggregates and reports the overall wallclock time in which the tasks in a system (or cgroup) wait for contended hardware resources.

The numbers here are percentages, not time itself:

The averages give the percentage of walltime in which one or more tasks are delayed on the runqueue while another task has the CPU. They’re recent averages over 10s, 1m, 5m windows, so you can tell short term trends from long term ones, similarly to the load average.

We can try to apply some I/O pressure by making a big tarball of a kernel source tree:

$ head /proc/pressure/*
==> /proc/pressure/cpu <==
some avg10=1.33 avg60=10.07 avg300=26.83 total=262389574

==> /proc/pressure/io <==
some avg10=40.53 avg60=13.27 avg300=3.46 total=20451978
full avg10=37.44 avg60=12.40 avg300=3.21 total=16659637

==> /proc/pressure/memory <==
some avg10=0.00 avg60=0.00 avg300=0.00 total=0
full avg10=0.00 avg60=0.00 avg300=0.00 total=0

The CPU is still under some stress here, but the I/O is now the limiting factor.

The output also shows a total= number, and that is explained in the patch as well:

The total= value gives the absolute stall time in microseconds. This allows detecting latency spikes that might be too short to sway the running averages. It also allows custom time averaging in case the 10s/1m/5m windows aren’t adequate for the usecase (or are too coarse with future hardware).

The total number can be helpful for machines that run for a long time, especially when you graph them and you monitor them for trends.

Running Home Assistant in a Docker container with a Z-Wave USB stick

The Home Assistant project provides a great open source way to get started with home automtion that can be entirely self-contained within your home. It already has plenty of integrations with external services, but it can also monitor Z-Wave devices at your home or office.

Here are my devices:

Install the Z-Wave stick

Start by plugging the Z-Stick into your Linux server. Run lsusb and it should appear in the list:

# lsusb | grep Z-Stick
Bus 003 Device 006: ID 0658:0200 Sigma Designs, Inc. Aeotec Z-Stick Gen5 (ZW090) - UZB

The system journal should also tell you which TTY is assigned to the USB stick (run journalctl --boot and search for ACM):

kernel: usb 3-3.2: USB disconnect, device number 4
kernel: usb 3-1: new full-speed USB device number 6 using xhci_hcd
kernel: usb 3-1: New USB device found, idVendor=0658, idProduct=0200, bcdDevice= 0.00
kernel: usb 3-1: New USB device strings: Mfr=0, Product=0, SerialNumber=0
kernel: cdc_acm 3-1:1.0: ttyACM0: USB ACM device
kernel: usbcore: registered new interface driver cdc_acm
kernel: cdc_acm: USB Abstract Control Model driver for USB modems and ISDN adapters

In my case, my device is /dev/ttyACM0. If you have other serial devices attached to your system, your Z-Stick may show up as ttyACM1 or ttyACM2.

Using Z-Wave in the Docker container

If you use docker-compose, simply add a devices section to your existing YAML file:

version: '2'
services:
  home-assistant:
    ports:
      - "8123:8123/tcp"
    network_mode: "host"
    devices:
      - /dev/ttyACM0
    volumes:
      - /etc/localtime:/etc/localtime:ro
      - /mnt/raid/hass/:/config:Z
    image: homeassistant/home-assistant
    restart: always

You can add the device to manual docker run commands by adding --device /dev/ttyACM0 to your existing command line.

Pairing

For this step, always refer to the instructions that came with your Z-Wave device since some require different pairing steps. In my case, I installed the battery, pressed the button inside the sensor, and paired the device:

  • Go to the Home Assistant web interface
  • Click Configuration on the left
  • Click Z-Wave on the right
  • Click Add Node and follow the steps on screen

Understanding how the sensor works

Now that the sensor has been added, we need to understand how it works. One of the entities the sensor provides is an alarm_level. It has two possible values:

  • 0: the sensor is tilted vertically (garage door is closed)
  • 255: the sensor is tilted horizontally (garage door is open)

If the sensor changes from 0 to 255, then someone opened the garage door. Closing the door would result in the sensor changing from 255 to 0.

Adding automation

Let’s add automation to let us know when the door is open:

  • Click Configuration on the left
  • Click Automation on the right
  • Click the plus (+) at the bottom right
  • Set a good name (like “Garage door open”)
  • Under triggers, look for Vision ZG8101 Garage Door Detector Alarm Level and select it
  • Set From to 0
  • Set To to 255
  • Leave the For spot empty

Now that we can detect the garage door being open, we need a notification action. I love PushBullet and I have an action set up for PushBullet notifications already. Here’s how to use an action:

  • Select Call Service for Action Type in the Actions section
  • Select a service to call when the trigger occurs
  • Service data should contain the json that contains the notification message and title

Here’s an example of my service data:

{
  "message": "Someone opened the garage door at home.",
  "title": "Garage door opened"
}

Press the orange and white save icon at the bottom right and you are ready to go! You can tilt the sensor in your hand to test it or attach it to your garage door and test it there.

If you want to know when the garage door is closed, follow the same steps above, but use 255 for From and 0 for To.

Allow a port range with firewalld

Managing iptables gets a lot easier with firewalld. You can manage rules for the IPv4 and IPv6 stacks using the same commands and it provides fine-grained controls for various “zones” of network sources and destinations.

Quick example

Here’s an example of allowing an arbitrary port (for netdata) through the firewall with iptables and firewalld on Fedora:

## iptables
iptables -A INPUT -j ACCEPT -p tcp --dport 19999
ip6tables -A INPUT -j ACCEPT -p tcp --dport 19999
service iptables save
service ip6tables save

## firewalld
firewall-cmd --add-port=19999/tcp --permanent

In this example, firewall-cmd allows us to allow a TCP port through the firewall with a much simpler interface and the change is made permanent with the --permanent argument.

You can always test a change with firewalld without making it permanent:

firewall-cmd --add-port=19999/tcp
## Do your testing to make sure everything works.
firewall-cmd --runtime-to-permanent

The --runtime-to-permanent argument tells firewalld to write the currently active firewall configuration to disk.

Adding a port range

I use mosh with most of my servers since it allows me to reconnect to an existing session from anywhere in the world and it makes higher latency connections less painful. Mosh requires a range of UDP ports (60000 to 61000) to be opened.

We can do that easily in firewalld:

firewall-cmd --add-port=60000-61000/udp --permanent

We can also see the rule it added to the firewall:

# iptables-save | grep 61000
-A IN_public_allow -p udp -m udp --dport 60000:61000 -m conntrack --ctstate NEW,UNTRACKED -j ACCEPT
# ip6tables-save | grep 61000
-A IN_public_allow -p udp -m udp --dport 60000:61000 -m conntrack --ctstate NEW,UNTRACKED -j ACCEPT

If you haven’t used firewalld yet, give it a try! There’s a lot more documentation on common use cases in the Fedora firewalld documentation.