May 15, 2026
Tiny DB, enormous comment-section energy
NanoTDB – Golang Append-Only Time Series DB
Tiny database for Raspberry Pi sparks big "isn't this just a log?" energy
TLDR: NanoTDB is a tiny file-based database for sensor-style data on small devices, built to be simple and crash-resistant. Commenters immediately turned it into a debate over whether it’s genuinely useful or basically just a fancy log file, with bonus dunks on Home Assistant’s data handling.
A teeny new project called NanoTDB is pitching itself as a small, built-in database for time-stamped sensor data on low-power gadgets like Raspberry Pi boxes, home sensors, and edge devices. The sales pitch is simple: everything lives in plain files, it writes data in order, and it’s designed to survive crashes without needing a bunch of extra software. In normal-human terms, it’s trying to be the tidy little notebook for devices that constantly record temperatures, pressures, and other readings all day long.
But the comments instantly turned this into a much juicier story: is this clever minimalism, or is it just a log file wearing a fake mustache? User xnx fired the opening shot with the brutally simple, devastatingly memeable: “How does this differ from a log file?” That one-liner basically became the mood. Then PunchyHamster arrived with the grizzled-engineer prophecy, joking that every append-only database starts life as a beautiful no-delete utopia before reality barges in and demands cleanup tools. It’s the kind of comment that says, “Congrats on your elegant design, see you in six months when users want an undo button.”
The most supportive reaction came from amluto, who dragged Home Assistant into the chat and declared its current time-series situation “comically poor.” Ouch. That turned the thread into a side quest about who actually needs this kind of lightweight tool and whether bigger names like ClickHouse or DuckDB already solve the problem better. So yes, NanoTDB launched as a tiny database announcement — and immediately got swallowed by the internet’s favorite sport: asking whether the new thing is secretly an old thing, but smaller.
Key Points
- •NanoTDB is described as a small embedded append-only time-series database for resource-constrained environments, storing all data in plain files with no external runtime dependencies.
- •Its architecture uses an engine that manages named databases, each with a write-ahead log, a metric catalog, and immutable partitioned data files containing compressed pages.
- •During ingestion, samples are parsed from line protocol, appended to the WAL for crash safety, added to an in-memory page, and later flushed as compressed page frames to append-only data files.
- •The system requires monotonically non-decreasing timestamps per metric and rejects out-of-order or stale samples.
- •The article details a compact WAL v2 format and a query path that scans partition headers to skip irrelevant frames before decompressing matching data.