Based on an open issue[0], it's suggested to run a server with 32GB+ of memory to handle hosting Clickhouse but that would mean self hosting Plausible would end up being $160 / month on DigitalOcean which would make it 10x more expensive than hosting my custom app that I want to see analytics for.
I know you can use less memory but it sounds like using less can result in an unpredictable environment where everything can stop working at any given moment depending on what Clickhouse wants to do. This happened to someone who replied in that issue. Their production set up stopped working because it ran out of memory.
Someone else wrote about it using close to 8GB of disk space to track ~8k page views at https://cyberhost.uk/plausible-3-month-review/. That was only written back in March 2021 too. They said they are going to look for an alternative solution because the the storage costs are too high.
The production environment that crashed due to Clickhouse OOM was our hosted product a while ago :) After that, we haven't had any downtime on our Clickhouse DB for over a year.
The issue with disk space stems from a bad default configuration. Clickhouse used to have EXTREMELY noisy debug level logging enabled by default with no rotation. This has been fixed in our hosting repo[1] so you get sensible defaults.
If you don't want to worry about downtime, planning disk space or compute capacity, then that's exactly what we offer at https://plausible.io. We process and keep the visitor data on our Hetzner servers in Germany.
[0]: https://render.com
Google has made sure that analytics for Google Ads works best within their own walled garden. Same with Facebook and Twitter with their Pixel products.
Instead of using the Referer header or utm parameters as intended, these large corps send obtuse random IDs (gclid, t.co/<id> links) which only they can correlate to an ad, search query or tweet using their internal database.
So until there is anti-trust action in this space towards more oppenness and competition, you're stuck with the ad provider if you want tight integration between ads and analytics.
