mirror of
https://github.com/scylladb/scylladb.git
synced 2026-04-23 10:00:35 +00:00
f1bbf705f9
This series extends sstable cleanup to resharding and other (offstrategy, major, and regular) compaction types so to: * cleanup uploaded sstables (#11933) * cleanup staging sstables after they are moved back to the main directory and become eligible for compaction (#9559) When perform_cleanup is called, all sstables are scanned, and those that require cleanup are marked as such, and are added for tracking to table_state::cleanup_sstable_set. They are removed from that set once released by compaction. Along with that sstables set, we keep the owned_ranges_ptr used by cleanup in the table_state to allow other compaction types (offstrategy, major, or regular) to cleanup those sstables that are marked as require_cleanup and that were skipped by cleanup compaction for either being in the maintenance set (requiring offstrategy compaction) or in staging. Resharding is using a more straightforward mechanism of passing the owned token ranges when resharding uploaded sstables and using it to detect sstable that require cleanup, now done as piggybacked on resharding compaction. Closes #12422 * github.com:scylladb/scylladb: table: discard_sstables: update_sstable_cleanup_state when deleting sstables compaction_manager: compact_sstables: retrieve owned ranges if required sstables: add a printer for shared_sstable compaction_manager: keep owned_ranges_ptr in compaction_state compaction_manager: perform_cleanup: keep sstables in compaction_state::sstables_requiring_cleanup compaction: refactor compaction_state out of compaction_manager compaction: refactor compaction_fwd.hh out of compaction_descriptor.hh compaction_manager: compacting_sstable_registration: keep a ref to the compaction_state compaction_manager: refactor get_candidates compaction_manager: get_candidates: mark as const table, compaction_manager: add requires_cleanup sstable_set: add for_each_sstable_until distributed_loader: reshard: update sstable cleanup state table, compaction_manager: add update_sstable_cleanup_state compaction_manager: needs_cleanup: delete unused schema param compaction_manager: perform_cleanup: disallow empty sorted_owened_ranges distributed_loader: reshard: consider sstables for cleanup distributed_loader: process_upload_dir: pass owned_ranges_ptr to reshard distributed_loader: reshard: add optional owned_ranges_ptr param distributed_loader: reshard: get a ref to table_state distributed_loader: reshard: capture creator by ref distributed_loader: reshard: reserve num_jobs buckets compaction: move owned ranges filtering to base class compaction: move owned_ranges into descriptor