cosven commented on code in PR #49383: URL: https://github.com/apache/doris/pull/49383#discussion_r2086886474
########## be/src/olap/base_tablet.cpp: ########## @@ -1768,6 +1768,148 @@ Status BaseTablet::update_delete_bitmap_without_lock( return Status::OK(); } +void BaseTablet::agg_delete_bitmap_for_stale_rowsets( + Version version, DeleteBitmapKeyRanges& remove_delete_bitmap_key_ranges) { + if (!config::enable_agg_and_remove_pre_rowsets_delete_bitmap) { + return; + } + if (!(keys_type() == UNIQUE_KEYS && enable_unique_key_merge_on_write())) { + return; + } + int64_t start_version = version.first; + int64_t end_version = version.second; + DCHECK(start_version < end_version) + << ". start_version: " << start_version << ", end_version: " << end_version; + // get pre rowsets + std::vector<RowsetSharedPtr> pre_rowsets {}; + { + std::shared_lock rdlock(_meta_lock); + for (const auto& it2 : _rs_version_map) { + if (it2.first.second < start_version) { + pre_rowsets.emplace_back(it2.second); + } + } + } + std::sort(pre_rowsets.begin(), pre_rowsets.end(), Rowset::comparator); + // do agg for pre rowsets + DeleteBitmapPtr new_delete_bitmap = std::make_shared<DeleteBitmap>(tablet_id()); + for (auto& rowset : pre_rowsets) { + for (uint32_t seg_id = 0; seg_id < rowset->num_segments(); ++seg_id) { + auto d = tablet_meta()->delete_bitmap().get_agg_without_cache( + {rowset->rowset_id(), seg_id, end_version}, start_version); + if (d->isEmpty()) { + continue; + } + VLOG_DEBUG << "agg delete bitmap for tablet_id=" << tablet_id() + << ", rowset_id=" << rowset->rowset_id() << ", seg_id=" << seg_id + << ", rowset_version=" << rowset->version().to_string() + << ". compaction start_version=" << start_version + << ", end_version=" << end_version << ", delete_bitmap=" << d->cardinality(); + DeleteBitmap::BitmapKey start_key {rowset->rowset_id(), seg_id, start_version}; + DeleteBitmap::BitmapKey end_key {rowset->rowset_id(), seg_id, end_version}; + new_delete_bitmap->set(end_key, *d); + remove_delete_bitmap_key_ranges.emplace_back(start_key, end_key); + } + } + tablet_meta()->delete_bitmap().merge(*new_delete_bitmap); +} + +void BaseTablet::check_agg_delete_bitmap_for_stale_rowsets(int64_t& useless_rowset_count, + int64_t& useless_rowset_version_count) { + std::set<RowsetId> rowset_ids; + std::set<int64_t> end_versions; + traverse_rowsets( + [&rowset_ids, &end_versions](const RowsetSharedPtr& rs) { + rowset_ids.emplace(rs->rowset_id()); + end_versions.emplace(rs->end_version()); + }, + true); + + std::set<RowsetId> useless_rowsets; + std::map<RowsetId, std::vector<int64_t>> useless_rowset_versions; + { + _tablet_meta->delete_bitmap().traverse_rowset_and_version( + // 0: rowset and rowset with version exists + // -1: rowset does not exist + // -2: rowset exist, rowset with version does not exist + [&](const RowsetId& rowset_id, int64_t version) { + if (rowset_ids.find(rowset_id) == rowset_ids.end()) { + useless_rowsets.emplace(rowset_id); + return -1; + } + if (end_versions.find(version) == end_versions.end()) { + if (useless_rowset_versions.find(rowset_id) == + useless_rowset_versions.end()) { + useless_rowset_versions[rowset_id] = {}; + } + useless_rowset_versions[rowset_id].emplace_back(version); + return -2; Review Comment:  确认一下这样理解对不对: 1. useless_rowsets 对应:无用 rowset 上的 delete bitmap 泄露。 2. useless_rowset_versions 对应:有用的 rowset 上的 delete bitmap 泄露。 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: commits-unsubscr...@doris.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@doris.apache.org For additional commands, e-mail: commits-h...@doris.apache.org