Key learnings from a disaster

Key learnings from a disaster
Photo by Marc Szeglat / Unsplash

Today, my peerverse.space was down because the disc ran out of space. In order to rectify the problem, I had to upgrade my storage. Because of that, I had to restore my instance from a back-up.

And that was rough. I lost a week's worth of data.

Part of the problem was that I was paying attention to the wrong metrics. According to Digital Ocean, I had 25 GB to play around with on my instance. My admin panel told me I was only storing appr. 100 MB of data. So I thought I was good to go indefinitely – this was not the case.

So what used up the data? It certainly wasn't my own posts. It was mostly data from the 250+ profiles I had been following on other instances. It took me a month to fill up 25 GB of data too.

Which tells me that the more people I follow on the Fediverse, the more of other peoples' data I host indefinitely unless I delete it.

As a small user instance, do I want to host other peoples' data in a place few people are going to see it twice?

If I want to be a good citizen of the Fediverse, maybe I do. There should be a record of data. I just don't know if anyone – other than me – will care about it here.

In a way, I wish Pixelfed allowed me to delete that data after a certain period of time because that would save me the effort of perpetually increasing my disk space.

Even for a small instance like mine, costs are going to increase the longer that the instance exists. And that's the problem for the long term viability of instances. Someone has to foot the bill, and inevitably it's going to be the admin.

This right here is how decentralization could inevitably result in re-centralization. Which instances are people going to trust? Ones that will stick around.

Which ones are going to stick around? The instances with enough capital to pay the bills. As with all things in our capitalist world, the "big" instances will eat the "small" instances.

Now understand, I'm not throwing mud at the Fediverse – just pointing out a flaw in its model.

For now, I'm going to continue with this test, because that's what it is. However, at some point I'm going to need to make a decision about where to go next.