I write about storage engines, consensus protocols, and the parts of distributed
infrastructure that are easy to draw on a whiteboard and hard to keep alive in
production.
Write-ahead logs, and why fsync lies
2026-11-02 · 12 min read
A storage engine is only as durable as its write-ahead log's fsync path. This note
walks through what the kernel actually does with an fsync, where battery-backed
caches sit, and the three places most production WALs silently drop the promise.
Tailing a replicated log without a leader
2026-10-19 · 9 min read
Reading from a follower is easy until you need to know that the follower is
caught up. Comparing counters, commit indexes, and lease-based reads — with
numbers from a small global cluster.
gRPC keepalives in long-lived connections
2026-09-28 · 7 min read
Default keepalive settings are tuned for callers that open and close connections
quickly. Batch pipelines do the opposite. What breaks, and which knobs matter.