Using Virt-Manager to Create Base VM Images

This topic is something that I've known about in theory, but never realized until the past couple of weeks ago how easy Virt-Manager (the KVM VM management GUI in Linux) makes this process. To anyone needing to test a lot of things in a Linux environment - or any environment really - and doesn't want to use containers, this will hopefully make your life a little bit happier. 

So I've been creating some self-install scripts for Nextcloud and other software - essentially ways to build up a basic Nextcloud instance from scratch with an easy-to-follow guided flow. In building the script if I hit a snag during testing, I would need to tear down the VM and start over. Thankfully, Virt-Manager can create VM clones. Even better it can leverage the base OS! 

 

If you uncheck the storage option, the image will boot just as the base OS image, and a diff partition will be created. This feature is great for testing, and while there is a notice running this way can be dangerous - essentially if the base OS ever changes the clone could get corrupted I imagine - for testing installs and different package combos, this is really nice. 

Essentially, this feature lets one install a base OS once. Create the user account, add sudo access, etc. once. Then shutdown the base OS, clone it with shared storage and start playing. No need to reinstall the OS, no need to worry if something goes sideways. If the test install somehow has issues, just nuke it, make another clone the good base and start over, or test something else. Really nice!

Why not just use containers?

Docker and podman and kubernetes, etc. are all fantastic tools. However, occasionally, having base feature sets like text editors, systemd, cron, etc, are really helpful. Especially if the instance will be running multiple services with different dependencies etc., using a VM makes a lot of sense. 

How to install it? 

I have a longer video way back in 2020 that goes through the concepts, but if using a Debian variety of Linux (Debian, Ubuntu, LinuxMint, PopOS, ZorinOS, MX Linux, etc.), just run the below command. For RedHat distros just swap 'apt' with 'dnf'.

sudo apt install qemu-kvm libvirt-daemon-system libvirt-clients bridge-utils virt-manager

Then start the libvirtd service

systemctl enable libvirtd

systemctl start libvirtd

That's all. Virt-Manager will be installed - search for Virtual Machine Manager - and start making VMs!

Using Nextcloud and WebDAV as a backup target

 

This video explores the built-in WebDAV feature of Nextcloud for backing up and sync'ing files from a local client to a Nextcloud instance. There are a couple of reasons why users or organizations may want to implement this feature.

- Nextcloud's own user management makes it very straightforward to separate our different user profiles, authentication, and also data quotas. This means it's very simple to deploy Nextcloud and automate employee data back ups on the network with automated back ups of say 30GB or 100GB per user.

- Using the WebDAV protocol can be more efficient than uploading lots of files to the web interface. 

- The Nextcloud desktop client also perform the same task, and also uses the WebDAV protocol. More information about the client tool is here: https://docs.nextcloud.com/desktop/latest/

 Additional resources:

• https://docs.nextcloud.com/server/latest/user_manual/en/files/access_webdav.html
• https://docs.qnap.com/operating-system/qts/4.5.x/en-us/GUID-31D5B05F-F29E-4D61-9758-C8CF839C14FD.html
• https://help.dreamhost.com/hc/en-us/articles/216473357-Accessing-WebDAV-with-Window

Object Storage on TrueNAS

 

I wanted to cover the many ways TrueNAS can be used to create and manage object storage. TrueNAS includes Minio S3 object storage out of the box, and the two work very well together. TrueNAS is a scalable, easy way to run and manage ZFS, and Minio is the de-facto way to self host S3 objects. 

Using the standard S3 service in TrueNAS is the quickest way to get setup and running with S3 objects. However, user management, if one wanted to open up remote access to the Minio web console, is a bit more difficult. 

What the video proposes is a couple of ways to silo off and breakdown the S3 service either with a web front end, or simply by making use of jails to host multiple, separate instances of the S3 object storage. 

Here, we'll look more at the details for each option. 

Create a jail to run Minio:

Go to the Jails section on TrueNAS. Create a new jail, give it a name, and set the network as desired - the video showed with DHCP, but static IP addresses are available as well. 

Once created Start the jail. Enter the shell. In the shell type 

pkg update

Accept yes to install.

pkg search minio

pkg install minio-0420....(whatever the current version is provided)

Once installed, make a directory. Could be anywhere, I chose in the /mnt directory of the jail.

mkdir /mnt/miniodata

Start the Minio server with the following

minio server /mnt/miniodata

That will start the service, but if you close the console/terminal screen the service will also terminate. To make this a bit more robust we can run the service with cron. 

Type crontab -e

Insert (defaul is vi, type 'i' to insert, esc to stop inserting, :w to write, :q to exit)

@reboot minio server /mnt/miniodata --console-address=":9090"

Now the service for Minio will start on each boot of the jail, with the console dedicated to run on port 9090 of the jails unique IP address.

Create a web server to run a frontend in a jail: 

To build out a  LAMP (technically a FAMP - FreeBSD, Apache, MySQL, PHP) I followed this excellent guide on Digital Ocean. I did not need the database portion, so that was skipped, though I did install the php-mysqli packages just in case I wanted it in the future. 

Digital Ocean Guide 

Install steps in the jail terminal/shell. 

pkg update -y

pkg install apache24

sysrc apache24_enable="YES"

service apache24 start

Navigate to the jail IP address and check to see if "It works!" appears.

pkg install php81 php81-mysqli mod-php81 php81-simplexml

The php81-simplexml package doesn't come down with the meta package for php81, and this threw me for a couple of hours because it is needed for the AWS PHP S3 plugin we install. 

Initiate PHP with specific settings:

cp /usr/local/etc/php.ini-production /usr/local/etc/php.ini 

Initiate PHP in Apache (you can use either vi or ee as the text editor):

ee /usr/local/etc/apache24/modules.d/001_mod-php.conf

Copy this in the file:

<IfModule dir_module>
    DirectoryIndex index.php index.html
    <FilesMatch "\.php$">
        SetHandler application/x-httpd-php
    </FilesMatch>
    <FilesMatch "\.phps$">
        SetHandler application/x-httpd-php-source
    </FilesMatch>
</IfModule>

Install PHP Composer for the S3 support.

pkg search composer

pkg install php81-composer 

composer require --working-dir=/usr/local/www/apache24 aws/aws-sdk-php

Now all this was to support the POC file I have over on Github. If you want to use it copy it to the jail, or install git in the jail and run a git clone command. Put everything in the /work/src/ folder in the /usr/local/www/apache24/data directory in the jail. Also be sure to modify the S3 endpoint, credentials, bucket and host address to those of the jail's IP and relevant credentials. 

I hope this gives a better overview of using object storage on TrueNAS. It is a really flexible feature, can be plugged into a lot of other environments, or even self-hosted on the TrueNAS itself using jails.

TrueNAS in 2023

 

I haven't had time to look back at TrueNAS in a while. This video is a raw install on a VM to explore the installation and setup of a basic SMB share. I had to scratch my brain a bit to remember some of the options to make some shortcuts. 

Some of the shortcuts that I took are well highlighted in the interface. 

• First of all I installed on a VM which is not recommended for production. It is known that ZFS doesn't like virtual media, but it is fine for testing and learning the system.
  
• Created a stripe pool - I only added one virtual disks because it doesn't matter for virtual medium - but it was nice the UI forces the user to confirm around 3 times before moving forward. 
• On a standard hardware installation, at least 3 drives - boot and 2 data drives, and more drives can be added to help scale capacity and performance.
• On the SMB share I didn't add a specific user, but rather just opened it up for anyone to modify. All the ACL flags, and allowing guest access is not great, obviously, for security.
  More information about creating a user to password protect the share is here:
  https://youtu.be/UEiwMIG0W9Q?feature=shared
  

I am planning to create some more videos about some of the lesser used features with TrueNAS as well as explore some of the newer features that have continued to improve over the past couple of years since I was using TrueNAS regularly.

Choosing a Laptop for Linux

Just sharing some experience I've had using various laptops with the latest Linux builds over the past 5 years.

In 2018 I started using Linux 100% of the time - literally no Windows or MacOS (never used it in the first place). I had a desktop, which I still have and still runs Linux, but I needed a laptop for work. 

I bought a then really cheap Acer S1 which had like a Pentium or Celeron processor and hard 4GB of RAM all soldered on. I thought it was fine as I could just use it for a customer presentation or typing up some things in a coffee shop. After a year, my wife wanted a PC so I gave that to her, and bought something new. Eventually she just didn't use it - preferring her iPad - so that got sold. It was okay up to like 8 browser tabs, but if I didn't have a desktop, I would have gone insane using it. 

As much as I love AMD, I've had issues. I had a Lenovo Ideapad for over a year and normally was great. Started with Linux Mint XFCE and really liked it. After a certain kernel update it didn't work - the screen was rendered completely unusable going from Kernel 5.0 to 5.3. Basically after reporting the issue and just booting to the old kernel each time, I got frustrated and abandoned ship to Fedora. That was great, but suddenly after about a year it started suddenly freezing randomly. After 6 weeks of this, I was writing an important email to a customer []7+complaining about something, and it froze. The laptop and the laptop stand suffered severe dents (punching laptops is not good for their health), and eventually I realized the second RAM slot (the only RAM slot that was removable and upgradeable) was rendered unusable - so only 4GB of onboard RAM to use.

That was donated to a school later....

I tried the Microsoft Surface Laptop Go for a while and it worked pretty well. The battery life on that is poor, but the heat management is bad on Linux, likely worse on Windows. I got a good deal so moved to it, but the performance was always hampered due to heat issues. This has turned into a Windows 11 PC for my wife who claims she needs it, I've not noticed using it.

A few months ago I bought a used Thinkpad X280 with an 8th gen Core i7, 16GB RAM, and and removable M.2 SSD. I messed up the screen - it's a poor 1360x768 px - but I normally connect to a larger monitor that's 1080P. This is the best laptop I've had. Build quality, keyboard, and just reliability. With PopOS it runs great, and Pop will even help show firmware updates for the system and install them with a reboot from the UI <-- something I never thought was possible in Linux. It cost less than USD300, and while it's maybe not the latest and greatest, it makes up for it by being reliable, good build quality, and strong Linux support.

Lessons learned. Older hardware is where Linux is developed, and is the better choice. Intel seems to be the better option long term - though this is not an absolute truth. Focus more on the build quality than the specs - unless you are building kernels, build quality will be more worthwhile than the exact Geekbench score for most use cases. Also consider heat management - make sure the model has vents, etc. Finally remember that Lenovo Ideapads are not Thinkpads <-- build quality, and general use by the Linux community (IBM invented Thinkpads and owns RedHat.....).

Connecting new Docker Containers to an Existing Docker Compose

Docker is amazing, but it has it's own way of handling things. 

Docker-Compose is amazing and it to needs to be used the right way. 

This post is about explaining how to attach new containers to existing containers already set up via docker-compose. 

Background:

Docker-Compose, when it brings up a bunch of containers or 'services' puts all of them in the same network with a common subnet and DNS so each service can talk to one-another based on the service name. 

Example:

mongodb://root:mongopwd@mongo:27017/

The 'mongo' in that line is the service name, and to connect say Mongo-Express to that container we can simply call the service name. Without the DNS component of the networking layer, we would need to know the specific IP of the container to connect. The 27017 is the port which is the default for MongoDB. 

Normally with Docker-Compose all the services or containers that run within the config file - normally docker-comopose.yml - will be given their own subnet and DNS to operate so they can just talk to one-another based on the service name. 

Example:

version: '3.1'

services:

  mongo:
    image: mongo
    container_name: mongodb
    restart: unless-stopped
    environment:
      MONGO_INITDB_ROOT_USERNAME: root
      MONGO_INITDB_ROOT_PASSWORD: mongopwd

  websvr:
    build: ./work
    container_name: web4mongo
    restart: unless-stopped
    volumes:
      - ./work/src:/var/www/html
    ports:
      - 80:80
    links:
      - mongo
    depends_on:
      - mongo

The services are denoted using indentation, and shown as 'mongo' and 'websvr'. Those two can than talk to one-another using 'mongo' or 'websvr' rather than the IP address of each individual container. This is really powerful since it makes the networking portion more easily understood, and also more easily repeatable since containers brought up on one system may get assigned different IP ranges than say on a different system. From a code perspective, just call the service name each time and all of that is handled. 

Adding another container:

However say we want to add a container. Often you need to adjust the docker-compose config, stop / remove the containers, and restart them. If the containers are removed and don't have a mapped volume to store the data already created, than the data could be lost. 

Less than ideal. 

However, with Docker you can map new containers to the network!

First find the network. Run 'docker network list'

 NETWORK ID     NAME                                DRIVER    SCOPE
50711*******   bridge                                bridge    local
ca9cc*******   host                                  host      local
118ef*******   lamp_default                          bridge    local
1887b*******   local-debug_default                   bridge    local
9299d*******   none                                  null      local
06023*******   ournoteorganizer_default              bridge    local
186fc*******   phps3object_default                   bridge    local
2209e*******   wordpress_default                     bridge    local

An output like the above may be what you see depending on the number of Docker instances you are running or have setup. This gives us the network names which we can use to attach a new container in the network. 

For this example we are adding a Mongo Express which is a web-based graphical UI that helps see and interact with a MongoDB database more easily. It's not necessary and should be disabled (docker stop mongo-express) when not used, but it is clearer, and possibly faster than running through all the mongosh commands.

Here is what we run:

sudo docker run --name=mongoex -p=8082:8081 \

--network=ournoteorganizer_default \

-e ME_CONFIG_MONGODB_ADMINUSERNAME=root \

-e ME_CONFIG_MONGODB_ADMINPASSWORD=mongopwd \
-e ME_CONFIG_MONGODB_URL=mongodb://root:mongopwd@mongo:27017/\

mongo-express

We create the new container using the run command, give it a name, define the port mapping between the host and the container, then assign it to the same network as the the original Docker-Compose group of containers - placing it in the same subnet and DNS / user space. Then we assign the necessary environment varilables to match the existing ones shown in the previous docker-compose.yml file. Finally specific the Docker image as mongo-express. 

This worked a treat for me since I had commented out Mongo Express in my pusblished OurNoteOrganizer application due to security concerns. Now I have a Mongo Express container I can call up when I need it or turn it off when I don't. 

Start: sudo docker start mongoex

Stop: sudo docker stop mongoex

Docker Compose is amazingly powerful, but understanding how it works is important to take full advantage of it. Knowing that it is essentially an automated networking stack + container creation makes it possible to work with without having to destroy containers and possibly lose information.

Ada & Zangemann

Ada & Zangemann is a book by Matthias Kirschner and Sandra Brandstatter that extols the benefits, fun, and importance of software freedom. A fun read for kids, it strives to instill a passion for inventing and tech into kids, and also has an important lesson for adults not to rely on any one person or organization for all their software needs. 

The premise of the story is around the juxtaposition of Ada - a bright girl without the means to afford many of the fancy tech most other families have - and Zangemann - the inventor of said tech. Through necessity, Ada tinkers with broken gadgets and old bicycles to give them a new life and make up for her lack of fancy skateboards and ice cream machines that other children have. In her drive to build the solutions she wants and needs, Ada finds her passion for problem solving, and hardware and software development. 

Zangemann, who is designed with an obvious reference to other high-tech oligarchs, is trapped by his own idiosyncrasies. He tends to go overboard in locking down devices to do only the things he wants to do - like his ice cream flavor of the day rather than letting people choose which flavor they want. He designs the products he likes, but wants them to be used in ways that only he deems appropriate. Obviously, the book is a bit over-the-top in some of the things Zangemann imposes on people, such as the iron that can't iron ties because Zangemann hates ties. That said many of the examples are probably more benign than the data and privacy society has given up to much of the tech giants of the world.

As the story progresses, Ada with her friends become more and more interested in tinkering with inventions and modding the many inventions that Zangemann sells. Zangemann becomes angry about the mods, and tries to get the government to step in and create laws to stop it. Ada and her friends, with the help of family protest, and over time end up not merely overturning the law, but also helping the government by creating new software without the limitations of the Zangemann tech. The government in the story saw the power Zangemann had over them with his digital monopoly, and our hero, Ada, was able to help out with her skills.

The story is a fun, inspiring adventure that I can relate to quite well. I want to be Ada! The political side is important as well, and speaks to the work Matthias Kirschner does as his day job as the President of the Free Software Foundation Europe (FSFE). This blog is really just designed to be a free resource to help people get off the ground and start learning about software, and honestly very little to none of the knowledge that I've learned would have been possible without free and open source software.

Ada & Zangemann is an approachable read that children and politicians alike can understand and hopefully take away some insight into the value of software freedom. Similar to this blog detailing my own experience learning about free tools and writing code, catching that feeling of "wow I built this" is incredibly important for young children. Free and open source code help enable the education, either in the classroom or self-taught, for those just getting started with technical development.

Free tools are resources that are essential for enabling more people to learn and develop new skills. It's very hard for many, myself included, to justify spending money on tools and software when I may not end up using it, or decide it's too difficult to learn, or I lose interest in something. However, if the tools are free, then one can use them and start to learn and develop one's skills further. If the tools are open source, then there is an even greater benefit because the code is more likely to have been peer reviewed, and more people can contribute so the project can become more permanent. Closed-source tools can always be stopped if the maintainer can no longer support the project. Open source is something that can always be maintained.

Ada & Zangemann has a political bent to it, but it is not wrong. Free software and open source are important and honestly need to be protected. This story helps spread that message in an approachable way, and I think is definitely worth a read for both kids and adults. After reading it with my son, he immediately started planning what he wanted to invent.

More info about the story can be found here: https://fsfe.org/activities/ada-zangemann/index.en.html

PHP with S3

 

This blog and video runs through getting started with PHP and S3. The subsequent code is all open source in the GitHub page located at the below link.

https://github.com/JoeMrCoffee/yourS3objects

Previously we looked at MinIO using it's console and UI, as well as quick set up in docker. This is a slightly more advanced look at interacting with object storage from a programmatic standpoint. 

PHP has an open SDK to tie in with S3 object storage that is provided by AWS. The video, as well as the source code in GitHub installs the SDK as part of the Docker bring up and build. This is done in the Dockerfile, and is very similar to installing the MongoDB packs required to make a PHP connection to a MongoDB database. 

From the code all the connection information to the S3 object store is in the 'header.php' file. If one wanted to use this site, or just its source code, they could also adjust the endpoint to the appropriate values. The 'header.php' file gets included in all the subsequent pages using the PHP include command, so everything else will follow the values in the header.php file.

The video talks a bit about the usefulness of the site. I developed it mostly as a reference for how the SDK could be used, but the actual site might be useful as a quick way for teams to just store and manage files in a particular bucket that only the admin or host could adjust.

Some more useful reference is below:

• MinIO documentation
• Linode AWS SDK for PHP
• PHP manual around file uploads - not about S3, but reinforces how the file upload process is handled with PHP

The 'sed' command in Linux

So I am building a new project involving PHP and S3 object storage, and as part of the solution I am looking at adjusting some of the defaults in the PHP configuration. This is not too hard to do with interactive commands, but a bit more complex when scripting the solution in a Dockerfile for automated install.

First hurdle: The docker image doesn't have a text editor. No nano, no vi, no nothing. This is actually a security measure. The general rule with containers is to make them as small as possible to keep the images small, but also improve security - the less on the server, the less can go wrong or get used the wrong way.

Second hurdle: Replacing a couple of lines of text in the php.ini file. 

Here is where 'sed' comes in.

When learning how to pipe or append data to a file I've often used | tee, or > or >> to send data to a file. 

Example:

$echo "some random text" > test.txt

That would create a file called test.txt with the 'some random text' as the content of the file. 

To append you can use the >> flag. Building on the previous example:

$echo "some more random text" >> test.txt

This adds to the file like below. 

some random text
some more random text

That is all fine, and works in the PHP containers from Docker Hub. However, if one uses the example php.ini files - php.ini-development or php.ini-production some of the values are written already. 

Here is where the 'sed' command can be employed. 

We have a very long config file with a bunch of comments and several settings that PHP can use to overwrite or adjust the defaults. In my case I want to change the default upload file size from 2M to something larger. With a text editor it is pretty easy to just search for that line and change it. To do this automatically, 'sed' is perfect. 

The syntax for substituting a line of text with sed is as below:

sed 's/<original text>/<replacement text>/' <file to change>

An example, if a file has the phrase 'some random text' we can change the sed output to 'some more text' with the below command.

sed 's/random/more/' filename

The output would be 'some more text'. 

IMPORTANT: 'sed' just changes the output, but it does not change the information in the file. To insert the changes to the file use the '-i' flag.

For example, if I wish to permanently replace the word 'random' with the word 'more', I would change the previous command to something like the following:

sed -i 's/random/more/' filename

This is a super useful command that honestly doesn't get a lot of attention, at least not in my decade plus of putzing around with Linux. Again really useful for scripting and I should have an example build in another couple of weeks.

A look at Snaps

 

Snaps are another way that developers package applications to make them installable in Linux.  Simlar to Flatpaks they are sandboxed and typically - though they don't have to be - used to run desktop applications on the Linux desktop. 

Snaps leverage the SquashFS file system which compresses the software for smaller footprints when installed on devices. When run the packages are uncompressed and mounted - shown in the video when running the 'df -alh' command. Because of the decompression needed when starting the application, sometimes the start up is a tad slower which is also shown in the video.

Snaps have had some controversy because the Snap store is hosted by Canonical and mirrors are not allowed. This organization is done for security, but there are members in the open source community who fundamentally resist this dynamic. Flatpaks would be the major alternative, though similarly, in practice, most flatpaks are hosted in a single instance in Flathub. 

More information about Snaps and contributing located below.

https://snapcraft.io/

https://en.wikipedia.org/wiki/Snap_(software)

Package managers: Flatpak - Linux Desktop Applications

 

Flatpak is quickly becoming the best way to get Linux desktop applications up and running on a Linux PC. Flatpak applications operate in a sandbox with their own images and environments that download and mirror the existing system in an isolated environment. Once installed, the applications themselves offer a responsive, essentially native experience.

Case in point, all my work requires me to use Google Chrome, Slack, and occasionally Signal on my Linux desktop. All of these can run as Flatpak applications, and in some cases perform even better or more reliably as a Flatpak install vs a native desktop client. 

When first installing Flatpaks, there is some overhead. When installing, if there is a specific environment - say a graphics driver / library - required, than the application will list it as an additional download. This can eat away at disk space versus a native application, but often it is trivial in the long run given the size of modern disk drives, and the fact that most of these environments will be re-used by other applications overtime. 

While not terribly in depth, the above video seeks to demonstrate quickly the different ways for getting started with Flatpaks, how to use the command line interface, and hopefully how to better use Flatpaks to get up an running with Linux on the desktop (i.e. a computer with point and click, not just a terminal).

More information:

https://flatpak.org/

https://docs.flatpak.org/en/latest/basic-concepts.html

Good info from System76

More specific permission settings that may be needed with certain Flatpaks

Turn down gamma in Linux under X11

 I purchased a second hand computer recently, an old Thinkpad X280, which is perfectly fine for most all of my needs. Only problem is the screen is awful. 

In an effort to slightly modify the contrast I found that there is a command in Linux under X11 (x-server is the traditional image renderer in Linux, so this may not work if your distro is using the new Wayland compositor). The command is xgamma. 

Example:

~$xgamma -gamma 0.7 

The above will reduce the gamma of the screen by 30% according to the X11 compositor. To make the changes stick, one can update the hidden file .profile in his/her home directory with the following:

xgamma -gamma 0.7 &> /dev/null

The &> /dev/null basically captures any would be output that results from the setting and throws it to /dev/null - basically throws it away. This is useful because otherwise there is an annoying "Okay to proceed" notice when first booting the system. 

Just another find which is available if not obvious in Linux! 

Reference:

https://linux.die.net/man/1/xgamma

https://forums.linuxmint.com/viewtopic.php?t=334249