Docker Compose has finally shipped native support for init containers, e.g. running DB migrations or similar one-off tasks.
Somewhat similar to k8s, they’re ephemeral containers that must exit successfully before the service containers start. They’re defined as pre_start hooks in the Compose file.
I shipped something very similar a few months ago for Uncloud designed as pre_deploy hooks.
There is no fundamental difference, only a few parameters differ. But Uncloud is a multi-host orchestrator where "deploy" is a distinct first-class operation so I believe the pre_deploy naming more clearly describes the intent to run something before the service rollout kicks in.
We may still adopt the new pre_start spec eventually, for example, to specify a container-lifecycle hook that runs before starting every service replica, not only once per service deploy.
How is AI adoption affecting team motivation?
https://t.co/BNUMQFabUM
> “TikTok-ification” of problem solving that turned engineering into “a babysitting chore, rather than doing interesting work.”
仕事がTikTok化しモチベーションが下がっている、という調査
Meta’s AI Storage Blueprint at Scale
https://t.co/wSHkZy3TDj
MetaがAIワークロードで発生するオブジェクトストレージのボトルネックを解消した話。メタデータ取得高速化・ブロックのあるインスタンスへの直接アクセス・(リージョン/インスタンス)キャッシュ
Dstl8
https://t.co/WymDLBojAP
Dstl8 investigates, correlates, reasons, and remembers runtime problems – across AWS, Kubernetes, OpenTelemetry, Supabase, and Vercel – to keep you out of debug rabbit holes. Context streams back into Claude Code, Cursor, and your dev flow
SQLens: Continuous Code-to-SQL Visibility in the Wild
https://t.co/0kK211Wtar
alibabaが社内で使っているsqlとコードを紐付けるやつ。IDE連携とかも。
LLMで推論するので実行前に紐付けられる。するとレビューでも使える
Finding a needle in a 4 GB haystack: from 0.75 GB/s to 49 GB/s in Go
https://t.co/uZRiNBFyVD
simdで並列 + ファイルを前から全部読むだけならmmapは効率が悪い(ページフォルトが多くなるから、大きいバッファで順に読むほうが速い)