Update: Managing docker containers with makefiles

I’ve been a bit of an old school “just use docker” vs. “docker-compose”, but it was the other day when I learned that they’ve now baked “compose” into the base docker cli. This means there is less reason to avoid using the features of docker compose.

Still, as previously discussed, I like using Makefiles to manage my containers. There is a very good argument to adopt using Watchtower, despite it being not recommended.

We do not endorse the use of Watchtower as a solution to automated updates of existing Docker containers. In fact we generally discourage automated updates.

Let me go on record that if you’re going to blindly run your update process, you may as well just use Watchtower. This is a good way to automatically break things of course, so maybe don’t auto-update things that you rely on to work (like wireguard or your web server).

With that out of the way – let me point out one of the downsides to my previous Makefile approach to managing containers: I’d slowly run out of disk space with many old images. I had gotten into the habit of running

every once in a while. This would get rid of all the containers, and images, that were no longer active. It did have the downside of removing images that were not currently running – so you need to be aware of that if you had assumptions.

Still my makefile approach saved me a few times allowing me to easily re-name / re-run the -old version of a container and back out of a bad update.

What I needed to do was make sure that the old, -old image was removed. This was where most of the disk space growth was. This turns out to be easier to do than to explain. Here is my update makefile

There are only two new lines. The first is

This is assigning the variable IMAGEID to have the docker image id value from the current -old image.

The second is the obvious removal of the image based on that id.

I believe that these two additional lines result in nearly no filesystem growth as I update images. Very handy as some of the linuxserver.io images pull down ~1GB of data every time, and they are updated weekly.

Debugging your network

The other day there was something strange happening on my home network. While my Amazon Fire HD8 tablet was happily on the WiFi, my Pixel 4a was ‘bouncing’ between WiFi and mobile data. The animated image at the top of the post is a simulation of what I saw.

The first thing I did was reboot my phone. Maybe this was some weird snafu and a reboot will fix it. This is the classic turn if off and on again.

The problem persisted. At the time I thought this may be just my device, but I found out later that Jenn’s Pixel 6a was doing the same thing, as was an older Pixel 2. The 6a was happy enough delivering internet, but based on the mobile data usage for the day it was clear most of the traffic was mobile data.

Next was the look at the OpenWRT router(s) – and I was a bit surprised that all of them had been up for 70+ days. I was headed out the door and decided to stall since it seemed that computers (and the 6a) were working ok (or so I thought). I confirmed that my phone was fine as I worked from the office, and had solid WiFi all day.

After dinner, it was clear that there was still a problem – but only the Pixel phones seemed to be having the problem. I tried removing the WiFi connection from my phone and re-adding it, maybe that would help? Nope. What if I switch to the guest network? Whaaat? Solid connectivity?!

Now the guest network doesn’t get the benefit of ad blocking from my Pi Hole. Maybe there is something going on there? I start poking around the logs on the Pi Hole, and reviewing logs on the OpenWRT router to see if there is any evidence. Nothing stands out. I try moving my phone to the IoT network – all three of the WiFi networks are hosted on the same hardware – so this helps eliminate the hardware and a good chunk of the software. The IoT network does make use of the Pi Hole, and to my surprise my phone was happy on the WiFi when using the IoT network.

At this point I’m about an hour and a half into debugging this, and I’m starting to run out of ideas. There have not been any configuration changes to my network recently. I don’t believe that any software updates have landed on my phone, and to have 3 different Pixel devices to all have the same problem all of a sudden is really weird. I’ve rebooted both my networking devices and the mobile devices – still the problem persists.

I run with 3 access points (two dumb AP and a main gateway), but each of them advertises the same SSID on two WiFi channels (a total of 6 distinct WiFi channels, all with the same SSID). This is a great setup for me as my devices just seamlessly move from connection to connection based on what is best.

My next idea is to change the SSID of one of the channels from ‘lan wifi’ to ‘hack wifi’ – allowing me to specifically connect to a given radio on a given access point. Now I can connect my phone to this new ‘hack wifi’ and know that configuration changes to this will affect the one device. Unsurprisingly the behaviour is the same, my phone just keeps connecting / reconnecting to this ‘hack wifi’.

I dive into some of the OpenWRT settings, looking for something that will make this WiFi connection more resilient. There are lots of options, but ultimately this is a dead end. I then wonder what would happen if I modify this WiFi connection to instead of routing to my ‘lan’ network, to route to the ‘iot’ network. Now my connection to ‘hack wifi’ works great.. hmm

What does this tell me? There seems to be something weird about my ‘lan’ network vs. there being something specific about the way that my ‘lan wifi’ is configured. This pivots me away from looking for differences in the WiFi configuration of my IoT, Guest, and Lan networks and looking more specifically at what’s connected to the lan network.

I grab the list of all devices on the lan network. Eliminate all of the WiFi clients in that list because it’s probably not them (but I’m guessing). Let’s take a closer look at the wired things (of which I have a good number). I start unplugging things, no joy. I turn off my main wired switch and still nothing. Finally I try unplugging my old server.

Boom. That was it. Almost immediately my phone connects to the WiFi network and stays connected. A quick check, and all of the other Pixel phones are happy now too.

This reminds me of a problem I had years ago, but my network was smaller and a little less complicated. A network cable I had built turned out to be bad, but only after months of use. Suddenly one day, my whole network was misbehaving, all devices were having problems. Powering off the machine had no effect, it was only once I removed the network cable that things came back to life. This was similar as the bad machine was connected to the main wired switch that I had powered off, but that wasn’t enough to fix the issue. Removing the cable was the fix.

Throughout this problem, computers on the ‘lan wifi’ seemed fine. Video calls worked fine, no strange drops or slow downs. Still, the impact to the Pixel phones was extreme.

The old server is very old (built in 2009), all of the drives in it are starting to fail and I should probably just power it off and wipe the drives and dispose of the hardware. I just haven’t quite gotten to it yet, this is probably a sign I should.

Edit  – Oct 6th

Oh no, it happened again. I haven’t plugged the machine or the cable that I determined was the issue last time back in – so what is triggering this?

This time I went straight to the 24port switch. Instead of powering it off (and on) which last time I did, and it had no effect – I removed it from the network by disconnecting the network cable between it and my OpenWRT router.

My phone immediately stopped bouncing between wifi and LTE, but just sat on LTE. It was then that I noticed that (of course) my pi hole which is run on a Raspberry Pi 4, is connected to the wired switch.

Plugging the switch back into the network, and my phone happily rejoined the network. The problem was gone, and has stayed gone since then. No reboots of any equipment were needed, only a brief drop from the network for the 24 port switch.  Very mysterious.

Edit – Oct 23rd

Ugh, again. This time I tried to be more surgical. I unplugged the pi-hole machine and waited 30 seconds. Plugging it back in might have fixed things? (not a power cycles, just a network drop?) Or maybe it fixed itself? Grr.

Edit – Nov 2nd

Yup, again. I woke up and noticed my phone was bouncing. Unplug the pi-hole from the network, wait, add it back.. magic – all fixed.

My ISP recently started giving me IPv6 (again) — I’m wondering if there is something funny going on there. This is still odd.

Edit – Dec 13th

And again (and possibly I missed recording one time it has happened since the last as well). It seems clear that disconnecting the network cable from the Pi and waiting a few (10) seconds – then reconnecting it is all I need to do to resolve this problem. I have changed (and tested) the network cable and changed the port it is connected to on the 24port switch.

Final Edit

I’ve failed to update a few times when I’ve had minor issues – but based on the history above you can see it continued. I’ve now moved the RPi4 (pi-hole) from connecting to the 24port switch, to directly connecting to my OpenWRT (Archer c7). Knock on wood, so far I haven’t seen the problem since.

Review: AKIYO O1 Mini Projector

I may completely lose my credibility as a home theater enthusiast, but hear me out. We are fortunate to have a cottage where we can get away from it all, and one of our rules there is “no tech”. Over the years this has eroded a little as we’ve ended up with generous mobile data plans, and our kids have become more technology attached. I also put the cottage on the internet for some automation setting up a remote site with openwrt and wireguard. Still, as a general rule we put our devices away and enjoy the sounds of nature and the freedom of being disconnected.

Under the guise of making things more fun for our teenage son, we decide to bring a movie to watch the other day. Watching Ratatouille on a 16″ laptop screen on the screened in porch after sunset was pretty nice. You can imagine the slippery slope we are on, and it was not a huge leap to think about setting up a projector and getting the ‘outdoor movie’ experience.

There are many great, low cost, projectors now. A local dealer has a great Epson 1080p projector for $700, and I’m certain with a little shopping we can cut that price down significantly – or look on the used market for a great deal at half the price. The Epson would be suitable for a ‘home theater’ space, and I don’t need my “no tech get away” to have an amazing audio video setup. This sent me looking at much less expensive solutions.

There is a whole class of low cost “mini” projectors out there, with so many choices. How do you pick one? You also have to wade through all of the technical mumbo-jumbo descriptions and marketing claims, many of them which are there to confuse you. There is also the misleading ‘sponsored’ reviews. The only good news here is there are many low cost choices so you’re not risking a lot, but remember the golden rule here: You get what you pay for.

I settled on the Akiyo O1 Mini Projector which I found on Amazon.ca. It had enough reviews which helps gives some confidence that they are not all sponsored/bought reviews. The price is well under $100 including taxes and delivery. If you bump your budget to $150 there are many more options, and many more features they promise. I decided that I was going to keep my outlay low and avoid too many features, I just needed a display.

This is what came in the box: The projector, a remote (2 AAA batteries not included), mini-tripod, power supply, HDMI cable and a couple of wooden q-tips (for cleaning?). The projector itself is very small, roughly 5.5″ x 4.5″ x 2.5″ – literally about two soda cans side by side. The HDMI cable is fairly long (5 feet?), but the power cable lead is quite short (3 feet?).

My initial impression was the fan was much quieter than I expected, and this was setup in my dedicated theater room which is nearly silent. The focus is pretty finicky, it feels like getting precise focus will be challenging as the control is quite loose. Keystone is limited, and appears to interact badly with focus. If you use keystone (which you probably will want) it becomes impossible to focus the top and bottom of the screen as the focus plane appears to be on an angle. It is all trade-offs here. This is still impressive for the cost.

Above is the start up screen (once you’ve done the initial setup) on my 80″x45″ screen (92″ diagonal). The projected image is larger than this screen by a good 8″ all around and there is ambient light in the room (dim, but not dark). Compared to my normal setup (native 1080p) the image feels soft, based on the image quality I’m starting to doubt if this is 720p.

Above is a screen shot of the projected image. Not bad eh? Image is fed to the project from my Macbook via HDMI. Now let’s look at the Macbook screen and the projected image together.

Ah, very different. This is a completely unfair comparison of course. While the projected image is watchable, the overall brightness and detail is lacking. The listed specifications for the Akiyo O1 are supposed to be 5500 lumens, but if there is any truth to this it’s the peak output of the light source not the amount of light coming out of the lens. I’ve seen guesses as to this being either 100 ANSI lumens, or 300 ANSI lumens. Maybe I’ll try measuring it one day, but it doesn’t matter for my needs.

I then moved the projector much closer to the screen, resulting in an approximately 45″ diagonal image. This made a huge difference, the on screen image now pops. Dark details are nice and visible now, and we still have good contrast. Viewing Ted Lasso, the colors were nice and bright and surprisingly good looking. No measurements here, and the human eye is very forgiving, but the feel of the image is impressive.

At this point in my notes, I call out that the fan noise is notably higher than my home theater projector. Certainly audible over the sound track of the movie during quiet parts.

My next steps were to put together something to feed the projector content. I tried just sticking one of my movies on USB stick and giving it to the projector, but this didn’t work. I decided while it was possible for me to figure out the right type of movie encode to feed the projector, having to re-encode any movie would be a pain. My choice here was to take a Raspberry Pi 3B I had sitting around and turn that into my media player.

To setup the Raspberry Pi, I visited the site and used the OSX specific installer. I took advantage of installer tool capability to pre-configure the install with a ssh server, my user and wifi information. The Pi booted and came up on my network with remote access, pretty slick.

I then installed mosh, one of my go-to utilities. To get sound working over HDMI I needed to run sudo raspi-config and navigate System->Sound->Output and change it from the headphone jack to HDMI. I also installed a screen keyboard: sudo apt-get install matchbox-keyboard, as I have a mouse plugged in but no keyboard. I finished up doing a system upgrade, just to make sure things are current.

Since VLC is built in to the base install, I was able to run that and play the movie just fine without having to worry about the format details.

The projected image is again around 45″ diagonal, and the photo was taken close to the projector itself making it look much larger. This isn’t a great reference image because it is fairly dark content and mostly blue/black. Still, a very watchable image – especially considering there was ambient light in the room too. Also, the wall being projected on is (mustard) yellow. The screenshot may not be selling you on this, but this convinced me I’d made a pretty good purchase. Fed via the Raspberry Pi I’m getting both sound and audio out, and the audio is quite loud enough. The image is bright enough to be engaging, and the focus isn’t that bad.

A couple more software tweaks. I wrote a simple bash script to watch for the USB drive to be mounted (so I could boot the Pi and then insert the movie later). Once the drive is detected, the first movie found will be played using VLC. I then used cron’s @reboot to run this script on start up. I discovered that while I could ssh into the machine and run the script and it would work, for some reason the cron invocation would lack sound. A simple work around was to setup key based ssh access, and from the script ssh to localhost to fix this, works fine this way.

I notice the Raspberry Pi was detecting the projector as a 1080p display and setting the screen up that way. I used the desktop GUI to force the HDMI input to 720p, this helped make the image sharper because we now have the native resolution matched. I believe this was the root problem with the ‘soft’ image I’d noted at first, with a proper 720p source things were pretty sharp.

The real test was taking a movie up to the cottage and having this all work there. We had the opportunity to try this soon after I had this figured out and it worked brilliantly.

The screen is a white bed sheet we pinned up, you can see the wrinkles and waves. The movie was Charlie and the Chocolate Factory. The displayed image was not anything close to videophile levels, but still I found myself sucked into the movie and deeply enjoying the experience. Yes, from time to time the fact that the image was distorted due to the waves in the sheet did distract a little. The fan noise was not an issue at all, the natural sound of the woods was louder than the fan, and the projector speakers plenty loud enough even at about half volume. It was overall a fantastic experience and easy to look past any of the imperfections, some of which are compromises in the setup vs. the technology.

When I consider my start in home theater, with expensive and large CRT projectors which had loud fans and could only resolve 720p images at best – this sub $100 unit with a long lasting LED light source is amazing value.