Uber just rebuilt its metrics database sharding to bound failure blast radius. m3db now places…
uber just rebuilt its metrics database sharding to bound failure blast radius. m3db now places shards in fixed-size subclusters, isolating replica groups and limiting the impact of a single node failure.
the fix is a scheduling problem, not a storage one.
Context
Uber Engineering's post of September 1, 2026, by Riya Tyagi, describes M3DB placing nodes into fixed-size subclusters, each owning a distinct non-overlapping slice of the shard space, with no shard shared across subclusters. Its example is 12 nodes at replication factor 3 with 6 nodes per subcluster, giving 2 subclusters. The baseline it describes is the sharded algorithm in which any node can own any shard, so one node failure can affect up to (n-1)/n of the cluster. Greedy shard donation keeps donor subclusters balanced when scaling.
The post gives constraints: the isolation group count must match the replication factor, instancesPerSubcluster must be a multiple of the replication factor, scaling happens in multiples of the subcluster size, and AddReplica is not supported. Its conclusion states benefits qualitatively, and no measured production numbers were found in the passages read, nor the rollout scope at Uber. The sharded algorithm remains appropriate for smaller clusters per the post. The note gives no date, and rebuilt reads as a placement algorithm option and not a storage redesign. The fix is a scheduling problem is the author's take.
Watch next
- Measured results and the M3DB release carrying the algorithm.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 22 September 2026 at 15:49 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →