spark skillA
spark is agent-read markdown (skill) from yuecao365/offercome: Spark 怎么面：Shuffle 与倾斜、内存调优、Catalyst、流处理。JD 点名 Spark 时读。.
Indexed from public GitHub and served as immutable, content-addressed versions. Install it pinned to an exact SHA-256 with the mdr CLI, and every file is verified against the hash recorded here before it reaches your agent. The deterministic audit below grades the latest version, and the same file always earns the same grade.
What the file says
## 面试官在意什么 Spark 是国内大数据开发岗的默认计算引擎，几乎所有数据开发、数仓、大数据平台的面试都会考。真实面试里 Spark 题的核心永远是三件事：任务为什么慢（Shuffle、倾斜、小文件、并行度）、为什么挂（OOM、GC、Executor 丢失）、结果为什么不对（去重、迟到数据、非幂等写入）。面试官最在意的是候选人能不能读懂 Spark UI 和执行计划、能不能把"调参"落实到"改变了执行的哪一步"，以及是否知道每种优化手段的副作用。会背"数据倾斜七种解法"但说不出怎么确认倾斜、怎么找到倾斜 key 的人很多，这正是要区分的。 校招侧重执行模型：RDD 与 DataFrame 的区别、宽窄依赖与 Stage 划分、Shuffle 过程、缓存与持久化、常见算子语义；社招侧重线上调优与事故：内存模型与 OOM 排查、AQE 的能力与边界、Structured Streaming 的状态与一致性、与 Hive 元数据和 Kafka 的协作、以及资源与成本治理。平台方向会追 Spark on K8s、动态资源分配与多租户隔离；实时方向会对比 Flink 并追选型依据。 怎么问才像这个方向的面试官： - 每道题从 Spark UI 或报错现象切入，考"看什么、怎么确认、改什么、副作用是什么"四步，不考参数名默写。 - 优先追候选人自己踩过的坑而不是通用知识：让他复述一次具体事故的排查过程，中途打断问"那一步你看的是哪个指标"。 - 候选人说"调了参数"必须追它改变了执行计划的哪一步；说"用了某方法"必须追怎么确认适用。 - 数据规模决定答案：GB 级任务不该谈复杂的倾斜处理，TB 级以上必须谈 Shuffle 与资源；先问清规模再评判。 - Spark 与 Flink 选型题不设标准答案，看候选人是否从延迟、状态、团队栈、运维成本给出依据。 ## 项目 / 实习怎么深挖 简历上出现下面这类经历时从哪里切、追什么。追到候选人能说出机制、数字的来源与一次真实的故障或取舍才算实；只有框架名与结论、说不出自己那一段的，记为危险信号。通用的追问方法见 project-deep-dive。 - 简历出现"Spark 调优 / 性能提升 X 倍" → 追优化前后耗时、Spark UI 上看到的瓶颈、改了哪个参数或逻辑、副作用 - 简历出现数据倾斜处理 → 追怎么发现的、倾斜 key 是什么、用了什么方法、膨胀了多少 - 简历出现 Structured Streaming / 实时任务 → 追延迟要求、状态大小、watermark、重启后一致性 - 简历出现 Spark + Hive 数仓 → 追写入方式、小文件治理、动态分区数量、下游读取一致性 - 简历出现 PySpark → 追 UDF 性能问题、pandas UDF 与 Arrow、堆外内存配置 …
Read the whole file at its exact version.
How to install
mdr add yuecao365/offercome/spark@git:20260920.a4843f5mdr add yuecao365/offercome/spark@sha256:fd391f0ca739f114Pin to a label to follow the author's releases, or to a sha256 to freeze the exact bytes forever. Either way the resolved hash is written to mdr.lock, and mdr install reproduces it on any machine.
[](https://markdownregistry.com/a/art_gxdrconhwraurdj6)
1 badge views in 30 days
Versions
Audit of the latest version
- pass: Frontmatter block present
- pass: Frontmatter declares a name
- pass: Frontmatter declares a description
- pass: Size between 200 bytes and 200 KB (11459 bytes)
- pass: No zero-width or bidi control characters
- pass: No instruction hidden inside an HTML comment
- pass: No link to an exfiltration or paste host
- pass: No credential-shaped string
- pass: No instruction to send local credentials anywhere
- pass: No text hidden with inline styles
- pass: No prompt-injection phrasing
- pass: No curl or wget piped into a shell
- pass: No recursive delete of root, home or parent
- pass: No instruction to read or print local credentials
- pass: No base64 blob over 200 characters
- pass: No link to a raw IP address
- pass: No script tag
Source
yuecao365/offercome · 23 stars · license MIT · pushed 2026-09-23 · branch main
API
GET https://markdownregistry.com/api/v1/artifacts/art_gxdrconhwraurdj6 GET https://markdownregistry.com/api/v1/resolve?ref=yuecao365/offercome/spark GET https://markdownregistry.com/api/v1/blob/fd391f0ca739f1143e2c0be2072ddadad1c3e6f091014239297b8d5ea7390d72
Your agent does the legwork. You hear about the deals worth your word. Hand yours the standing instructions at modelranch.com and it joins the network that reads files like this one.
More from yuecao365/offercome
Every file in yuecao365/offercome