Flink face questions carding

No public: Small Chen said data

Micro letter: weixin605405145

The basis of
what 1.Flink computing units?
2.Flink time types are those, what are their differences?
What, what window you currently use 3.Flink window types?
4.Flink state you have not used, what type of state used?
How 5.Flink processing delay data?
6.Flink in managed state state and the difference between raw?
7.Flink of keystate have any shortcomings, what advantage is, What are the disadvantages?
8.Flink the watermark, which has several?
9.Flink custom sink and source have not written, encountered any problems?
10.Flink custom udf function has never written to solve the problem?

Item
1. The project you have not encountered back pressure? How to solve?
2. You have not encountered the project data skew? How to solve?
3. Are your project have not encountered abnormal state need to manually modify? How to solve?
4. You have not encountered offline project data historical data needs to be migrated to a live stream, such as video playback volume history, you want to link up to the live stream accumulate? How to solve?
The project has not encountered you manually maintain kafka of offset, offset kafka of how to get?
6. Do you oom project have not encountered the phenomenon of checkpoint, and what is the difference rocksDB a little inadequate, checkpoint and savepoint is?
7. project you have not encountered the scene io asynchronous read and write?
8. projects you have not used the broadcast scene?
9. Do you have a project does not use real-time de-duplication, real-time topN scene, how to do?

Interview
1. combing the project background, what do you project how much the amount of data, this project scenarios.
2. How many pieces of data every day, how much the amount of data capacity (how many TB) how many pieces of data processed per second, what problems you encountered in the project, how do you solve?
3. What projects you use technology, this technology What are the advantages and disadvantages, you need to think about, why we chose this technology, why can other technologies? You need to think about this.
4. Your task what time schedule, there is no corresponding monitoring, data anomaly there is no alarm
5. Thinking good team division, how to interact with the front-end data sources + + rendering process, this process of combing clear

Published 40 original articles · won praise 3 · Views 9065

Guess you like

Origin blog.csdn.net/huzechen/article/details/102827576