Typical Interview Questions for Offline Data Warehouse Positions (Spark, Hadoop, etc.)

The article's protagonist is a fresh graduate looking for an offline data warehouse summer internship, having interviewed at companies such as Ctrip, Meituan, and Alibaba, with roughly ten interviews and ultimately receiving a satisfactory offer.

The typical interview pattern is: based on the resume description, the interviewer asks a broad question to see how widely and deeply you can speak, then follows up with deeper probing on specific knowledge points, often digging down to source‑code implementation level.

Project experience occupies a large portion of the interview and is closely tied to the resume projects. Other interview questions include:

Alibaba

1. 和 MR相比 spark 的劣势在什么地方？<br/>2. spark内存溢出。为什么会内存溢出，如何解决？<br/>3. spark shuffle 的原理，如何解决数据倾斜问题。（还是围绕内存溢出）<br/>4. 离线数仓，结合项目某个业务领域：<br/> 4.1. 考察建模理论，结合实际应用。<br/> 4.2. 考察数仓建设的流程。<br/>5. 数据质量相关（保证数据能够准时的出来，不出错，效率高）<br/>6. 谈谈对 实时仓库的理解（Flink，解决了什么问题和需求）<br/>7. Spark和MR的区别<br/>8. Spark和Hive的区别<br/>9. 聊数仓，对数仓体系结构的了解，最令人印象深刻的是哪一层的建设。<br/>10. 自己的科研项目，如何应用到实际的工作场景中。<br/>

Meituan

1. spark RDD的理解。<br/>2. spark 内存的情况。<br/>3. spark OOM的情况，如何解决。<br/>4. spark 数据倾斜，如何处理。<br/>5. Yarn如何实现资源分配和调度的？Yarn如何保证调度的可靠性。<br/>6. HDFS如何进行读写。<br/>7. 拉链表是什么，有什么作用，用在哪里，优缺点有哪些。<br/>8. 一个java程序运行起来，会发生什么事情。<br/>9. 浏览器中输入一个网址，按下回车，发生哪些事情。<br/>10. SparkSQL遇到慢查询如何处理<br/>

Ctrip

1. 数据倾斜如何解决。<br/>2. 拉链表如何使用。<br/>3. ChatGPT对数仓的影响<br/>

At the end, the article includes a promotional link to a large‑scale big‑data interview community.