Data Engineer

What additional parameters of DDL/CREATE TABLE in ClickHouse do you know, what do they affect, and in what cases are they used?

How to distribute storefront data unevenly across three shards: 50% on the first and 25% on each of the remaining two? How to implement this?

How were tables created on the cluster, and how was the problem of missing replicas on a shard due to ZooKeeper solved?

How did you solve optimization problems when a query runs slowly and performs a full scan? How would you approach a new such task?

How exactly did you read Parquet?

In what cases can a Materialized View in ClickHouse miss data or duplicate it?

Have you encountered the ClickHouse 'Too many parts' error during insertion? How did you resolve it?

What Python libraries have you worked with?

Tell me more about data skew between shards: how was it determined and how was the corresponding function/mechanism used?

Are you familiar with UDF (User Defined Functions) in ClickHouse, have you worked with them?

How did you fix data skew by client_id when one client was significantly larger than others? What options are available?

What parameters of the Distributed engine in ClickHouse are you familiar with?

How would you handle a Materialized View chain through an intermediate ReplacingMergeTree table to correctly populate data that already exists before launching a new MV?

Which version of Airflow have you worked with?

On which fields did the data merging occur and how were duplicates eliminated in the view?

There is a large PostgreSQL table into which new records with a monotonically increasing primary key are constantly arriving. How to organize data loading into ClickHouse?

Are you familiar with ReplicatedQueue? Have you worked with it?