Fonts look stupid, uptime history (in terms of what is shown on the status page) is limited to like 30 days or something, theming is not perfect
No, but it works great for my log volume which is like 3 GB/mo. They're the only solution I found with reasonable resource consumption and a simple software architecture (just sqlite and a rust backend), everything else is like "you have to have clickhouse as a dependency plus 3 other things" -> multiple gigs of ram eaten up immediately for no reason
In terms of features, does it make you happy - too many, not enough, or just right?
ClickHouse doesn’t really consume much memory, but I hear you. I built my error tracker on the same stack (SQLite + Rails). Rust is probably friendlier toward memory than Ruby, though.
I’d call consuming multiple gigs of ram while zero logs are being ingested (system at idle) way too much.
Openobserve eats an order of magnitude less RAM and cpu at idle. Ultimately higher resource consumption leads to higher power bills so it’s something I pay close attention to (I own my own hardware)
I’ve used ClickHouse on the cheapest Hetzner server with an app that was never in production (basically a test environment) under artificial load, and I didn’t have any OOM errors.
Plausible also uses it and did have some hiccups, so I chose one tier above the cheapest.
Have you heard of Kamal? It makes deployments on your own hardware a breeze.
I didn't have OOM errors, but it's still more RAM usage than I'm going to tolerate. No reason it should eat multiple gigs doing absolutely nothing. I reserve as much RAM as possible for my own apps.
I tried Kamal some time ago and it was very buggy with terrible documentation. I think it improved but I don't care since I already started using Dokku which works fine for me.
Fonts look stupid, uptime history (in terms of what is shown on the status page) is limited to like 30 days or something, theming is not perfect
No, but it works great for my log volume which is like 3 GB/mo. They're the only solution I found with reasonable resource consumption and a simple software architecture (just sqlite and a rust backend), everything else is like "you have to have clickhouse as a dependency plus 3 other things" -> multiple gigs of ram eaten up immediately for no reason
Good enough
I’d call consuming multiple gigs of ram while zero logs are being ingested (system at idle) way too much.
Openobserve eats an order of magnitude less RAM and cpu at idle. Ultimately higher resource consumption leads to higher power bills so it’s something I pay close attention to (I own my own hardware)
I’ve used ClickHouse on the cheapest Hetzner server with an app that was never in production (basically a test environment) under artificial load, and I didn’t have any OOM errors.
Plausible also uses it and did have some hiccups, so I chose one tier above the cheapest.
Have you heard of Kamal? It makes deployments on your own hardware a breeze.
I didn't have OOM errors, but it's still more RAM usage than I'm going to tolerate. No reason it should eat multiple gigs doing absolutely nothing. I reserve as much RAM as possible for my own apps.
I tried Kamal some time ago and it was very buggy with terrible documentation. I think it improved but I don't care since I already started using Dokku which works fine for me.