Dispatches / a short note
Operating clickhouse at petabyte scale
He had operated petabyte-scale ClickHouse clusters for 5 years
Mostly about choosing the right storage (local / cloud), sharding vs replica with partitioning - based on latency or QPS requirement. Oh I forgot about choosing compression algorithm, problems during upgrades, injestion side-effects and handling backpressure. I could add part 2 finding here too (but more on this later).
https://www.tinybird.co/blog/what-i-learned-operating-clickhouse
https://www.tinybird.co/blog/what-i-learned-operating-clickhouse-part-ii
Discussions:
https://news.ycombinator.com/item?id=49601138