A relaxed read waits only for a lower bound: the no-op entry the leader wrote when it was elected. That entry is almost always applied already, so a relaxed read almost never waits, and it is still linearizable.
12
949
drdr.xp retweeted
New Claude Sonnet 5.5 eval results are in! xhigh shines this time, while max makes you wait.
We ran Sonnet 5.5 vs. Sonnet 5 on our in-house OpenResty Coding Evals and our harder Bug-Hunting Evals, at both xhigh and max effort.
Coding Evals
• xhigh: quality 77.6% → 87.0%, avg time per task 9.4 → 9.7 min (roughly flat)
• max: quality 76.4% → 88.6%, avg time per task 26.0 → 49.5+ min (lower bound shown in chart)
Bug-Hunting Evals
• xhigh: quality 76.9% → 87.7%, avg time per task 33.1 → 18.0 min. Better AND faster!
• max: quality 79.4% → 86.1%, but avg time per task 30.5 → 109.9 min (~3.6x slower)
At least on our evals, Sonnet 5.5 is a clear quality upgrade. xhigh strikes a better balance: no slowdown on writing new code, and much faster at bug hunting. max takes way longer, and this time it even scored slightly below xhigh on bug hunting… Picking the right effort level really depends on the task.
OpenRaft 0.10.0 is released.
OpenRaft 0.10 makes most of its types pluggable, moves optional parts into separate crates, and adds configurable quorum rules, pipelined replication, Pre-Vote, leadership transfer and faster linearizable reads. It also fixes several safety bugs.
What's new:
• Configurable types: RaftTypeConfig selects the term, leader id, vote, payload, membership, error source and client-write responder.
• Quorum rules: a membership type defines separate read and write quorums and per-node roles. The new openraft-membership crate provides threshold, weighted, grid, hierarchical and explicit memberships. A membership change writes one log entry when the old and new quorums intersect.
• Replication: the leader pipelines AppendEntries through RaftNetworkV2::stream_append(). Followers send conflict hints that shorten log matching, and consecutive appends merge into one storage write.
• Elections: optional Pre-Vote stops partitioned nodes from inflating the term, and Trigger::transfer_leader() hands leadership to a chosen node.
• Linearizable reads: each read chooses its own policy, one heartbeat round answers every waiting read, and lease reads skip the RaftCore round trip.
• New crates: runtimes for tokio, compio, monoio and deterministic simulation; openraft-multi shares connections across Raft groups; openraft-metrics-otel exports OpenTelemetry metrics; openraft-legacy keeps the v1 network API.
• Safety fixes: a restarted leader no longer reuses log ids for different entries, a leader commits a blank entry of its own term before it changes membership, and a restarted follower keeps its vote lease, so a lease read on the old leader cannot miss a committed write.
Upgrading from 0.9 requires code changes, and the upgrade guide covers each one. With the default types, data stored by 0.9 needs no migration. OpenRaft 0.10 needs Rust 1.88 or later. The 0.9 branch keeps receiving bug fixes until 0.11 development starts.
Thanks to all the contributors to OpenRaft.
Release notes: github.com/databendlabs/open…
Upgrade guide: docs.rs/openraft/0.10.0/open…
11
534
drdr.xp retweeted
New in Databend Cloud Fundamentals: Episode 06. 🧐
Learn how Streams capture incremental table changes, when to use Append-Only vs. Standard mode, and how consumed records make way for the next batch.
Watch 👉 youtu.be/2G7yGuGFhqY
OpenRaft v0.10.0-alpha.38 is out:
• Custom quorums: weighted, hierarchical, grid & more
• Single-entry membership changes when every old quorum intersects every new quorum
• Local lease reads skip the RaftCore round trip
Breaking APIs; upgrade notes:
github.com/databendlabs/open…
6
380
OpenRaft now lets an application define its membership.
A 4-node cluster elects with 3 votes and commits after 2 copies.
Built-in types cover node count, weight, sites, and grids. You can also define your own.
blog.openacid.com/distribute…
6
408
drdr.xp retweeted
OpenTelemetry + Databend Cloud: logs, traces, and metrics in SQL tables. 🚀
See how the Collector exporter works, with a payment-failure demo showing how to join logs and traces to analyze order impact.
Read the walkthrough 👉 bit.ly/4eir49M
drdr.xp retweeted
New in Databend Cloud Fundamentals: Episode 05. 🚀
Load data from Stages, S3, remote servers, and local files with COPY INTO—then transform it during ingestion and handle evolving schemas.
Watch ▶️ youtu.be/OdazahbtMnA
A minor optimization for single step Raft config change: github.com/drmingdrmer/conse…
2
7
672
drdr.xp retweeted
New in Databend Cloud Fundamentals: Episode 04. 🎉
Learn how User, Internal, and External Stages work—then create, query, and manage Stages, Databases, and Tables with SQL.
Watch 👉 youtu.be/1hwBBLGCP_g
#Lakehouse #DataEngineering
OpenRaft now skips joint consensus when every old quorum intersects every new quorum blog.openacid.com/distribute…
Raft Membership Changes: When a Leader Must Commit a Current-Term Entry First
blog.openacid.com/distribute…
4
382
I think the best use of AI is reducing work entropy, across docs, codebases and writing. In other word: make it short. Condense the content so information stays easy to grasp
1
1
469
The 2nd step of Raft’s joint config change `c01 → c1` does not require the leader’s first log to be committed(single step change requires it, 1st step of joint doesn't) — unless bidirectional transitions are permitted, meaning both `c01 → c0` and `c01 → c1` are allowed.
1
299
drdr.xp retweeted
272 shipped changes across Databend’s August–September GitHub Releases. 🚀
Highlights: materialized views, cross-tenant Data Sharing, SQL-queryable lineage, Fuse partitioning, TopN pruning, and reliability fixes.
Read the roundup 👉 bit.ly/47eD5sW























