DigitalNigel

For the last several years I’ve had a storage shelf attached to my home server supporting roughly 30TB of redundant storage making use of ZFS, which I’ve been quite pleased with. I use it to store all my various media, and back up all of my other computers to it. As part of my backup scheme I also back up some of the data from the array to a remote array in another location.

Unfortunately as the years have passed the data stored has grown and the array has been nearing full for quite some time. I’ve put off expanding the array as there’s always more pressing/interesting things to spend project money on, and a bit of a battle with myself about *really* needing more than 30TB of storage.

Finally a deal came along on disks that was too good to resist, and I ordered twelve more 6TB disks to complement the eight I had already. As is always the way it goes, this also necessitated changing out the storage shelf I had to accommodate the additional disks as well. Fortunately a friend was swapping some of their storage hardware at the same time and had some HP D2600 storage shelves they gave to me for a good price.

Racking up two of them, and plugging them in to my HBA was pretty quick and painless. After which comes installing all the drives, and configuring the ZFS pools.

With only a minor hiccup, in that ZFS needs the same level of redundancy for each sub-pool, I couldn’t add the new twelve disks as a new sub-pool to the existing eight. I needed to create a new, separate, ten disk pool, move the data over, destroy the old pool, and create a new sub-pool with the original eight plus the two left over from the twelve. With all that done I’ve now got a bit more than 80TB usable space and plenty of room for growth so I can forget about it for the next few years.

Some friends and I recently spoke about understanding our home energy usage better, and implementing data collection for various circuits in which we had little real visibility into the actual usage and how they contributed to our energy bills.

We found an open source project making metering boards that would handle the sensing, provided with external clamp-on current transformers. Some searching, and a group bulk order later, we had a sensor board stack based on the ESP32 and a bunch of transformer cores to start working with.

We found a library for CircuitPython that had the start of what we needed, but appeared to be abandoned for some time, and needed some additional help to do everything we needed it to. After reaching out to the prior creator, we forked the repo and have been making a fair number of improvements/additions.

The new code is working well, and I have it tied into my general monitoring system, providing nice graphs. Here are some example graphs, with a longer timescale one covering several days, where we can see the heat pump usage as it works harder during the night as temperatures fall, and usage drops during the day as the temperatures come up. As well as a shorter timescale graph covering just 24 hours were we can see the individual cycles the system is running that the graphing system averages out over the longer timescale.

I have a multi-zone ductless mini-split system installed in my house, which has been a fantastic addition and does a great job with keeping the house comfortable. Though the limitation of all zones having to be in the same mode because the compressor unit outside can only do one mode at a time.

This leaves a coordination task, as temperature set points change through the day to save energy, and mode switches move between heating and cooling as the weather changes, particularly in the spring/fall transition seasons when daytime temps could require cooling, and nights still need heat.

It’s certainly possible to manage this with the individual remotes for each zone, but it can get to be annoying to tweak all the zones a couple times a day. I decided to build a ESP32 based IR controller with a central control and schedule to ease the task.

I also took it as an opportunity to experiment with some small E-Ink displays, which are well suited to long term, low power, battery operation for the remotes.

As I was building a half dozen devices, it was a great choice for assembly with my pick and place machine I’ve written about before.

Finishing out the assembly, attaching the display and battery, and 3D printing a case for the unit wraps the project into a neat little package.

Each unit sleeps at low power using only 10uA, waking up every 15 minutes to report sensor data, check the schedule with the control system, and if scheduled send an IR command emulating the factory remotes. Overall the 2000mAh battery lasts about three months between a recharge.

The factory remote protocols were reverse engineered using my Saleae Logic analyzer, and when available docs others have written about the same or similar devices.