ClickHouse Coordinates a Cluster-Wide Backup with ZooKeeper
A cluster-wide backup is a ZooKeeper-coordinated election: one owner per data file, a stage barrier between phases, and a consensus loop over metadata.
Posts
All posts published on this site.
Everything published here, newest first.
ClickHouse Deletes Expired TTL Rows Only When Parts Merge
ClickHouse stores a per-part TTL expiry bound and removes expired rows during a merge, using a bit mask rather than a rewrite.
ClickHouse Generates an Unbounded Number Stream in Constant Memory
numbers() is a generator that iota-fills one column per block and keeps only a counter in memory.
ClickHouse Incremental Backups Store Only the Changed Tail of a File
Incremental backups reuse the base backup’s bytes: only the suffix after a verified shared prefix is uploaded, and the restore concatenates the two halves.
ClickHouse Keeper Carries a Tree Digest in Every Raft Log Entry
Keeper puts a digest of its node tree inside every Raft log entry; a follower whose recomputed digest disagrees with the leader’s kills itself.