Prev Next

BigData / Apache Iceberg Interview questions

What is a manifest file?

A manifest file is an Avro file that lists the actual data files belonging to a specific portion of a snapshot, with one record per data file, including that file's physical path, format, and per-column statistics like value counts, null counts, and min/max bounds.

Because a manifest carries min/max bounds for every column of every data file it references, a query engine can use those bounds to skip reading an entire data file outright if the file's statistics prove it can't possibly contain any rows matching the query's filter — a form of file-level pruning that happens before any actual data reading occurs, purely from metadata.

Iceberg reuses unchanged manifest files across snapshots wherever possible — if a write only touches a specific partition, manifests describing untouched partitions don't need to be rewritten — which keeps the metadata overhead of frequent small commits manageable rather than requiring a full metadata rewrite on every single write.

What does a manifest file list, with one record per entry?
What lets a query engine skip reading a data file entirely, based purely on metadata?

More Related questions...

What is Apache Iceberg? What is the purpose of Apache Iceberg? What is a table format, and how does it differ from a file format? What are the key features of Apache Iceberg? What is the architecture of an Iceberg table? What is a snapshot in Apache Iceberg? What is a manifest file? What is a manifest list? What is the table metadata file? What is an Iceberg catalog? What is hidden partitioning? What are partition transforms in Iceberg? What is schema evolution in Iceberg? What is time travel in Apache Iceberg? Which query engines support Apache Iceberg? What file formats does Iceberg use to store data? How do you create an Iceberg table? What is the difference between Iceberg and a Hive table? What are field IDs in Iceberg, and why do they matter? What is ACID compliance in the context of Apache Iceberg? What is the difference between Apache Iceberg and Delta Lake? What is the difference between Apache Iceberg and Apache Hudi? Explain how hidden partitioning differs from Hive-style partitioning? What is a lakehouse, and how does Iceberg enable it? What is partition evolution, and how does it work internally? What is the difference between copy-on-write and merge-on-read in Iceberg? What are positional deletes versus equality deletes? Explain the internal working of Iceberg's snapshot isolation mechanism? What is the REST catalog, and why has it become important? What is the difference between a Hive catalog and a REST catalog? Explain the execution flow of a query against an Iceberg table? How does Iceberg achieve schema evolution without rewriting data? Explain the internal working of manifest-level partition pruning? What is compaction in Iceberg, and why is it needed? How do you perform time travel queries in Iceberg? What is the difference between a snapshot rollback and time travel? Explain how Iceberg handles concurrent writes? What is the role of sequence numbers in Iceberg snapshots? Explain the lifecycle of a write operation (commit) in Apache Iceberg? What are branches and tags in Apache Iceberg? How does Iceberg support upserts via MERGE INTO? What is the small file problem, and how does Iceberg address it? Explain the internal working of column-level statistics in manifest files? What is the difference between Iceberg V1, V2, and V3 table specs? How do you migrate an existing Hive table to Iceberg? What are deletion vectors, and how do they improve on positional delete files? Explain how Iceberg integrates with Apache Spark for reading and writing? What is metadata table querying in Iceberg? How do you troubleshoot slow query planning on a large Iceberg table? Explain the execution flow of building a streaming lakehouse pipeline with Iceberg and Flink?
Show more question and Answers...

Web

Comments & Discussions