Fix regression issue of too large score (#10518)

Summary:
https://github.com/facebook/rocksdb/pull/10057 caused a regression bug: since the base level size is not adjusted based on L0 size anymore, L0 score might become very large. This makes compaction heavily favor L0->L1 compaction against L1->L2 compaction, and cause in some cases, data stuck in L1 without being moved down. We fix calculating a score of L0 by size(L0)/size(L1) in the case where L0 is large..

Pull Request resolved: https://github.com/facebook/rocksdb/pull/10518

Test Plan: run db_bench against data on tmpfs and watch the behavior of data stuck in L1 goes away.

Reviewed By: ajkr

Differential Revision: D38603145

fbshipit-source-id: 4949e52dc28b54aacfe08417c6e6cc7e40a27225
main
sdong 2 years ago committed by Facebook GitHub Bot
parent f3ddbe66bd
commit 2297769b38
  1. 53
      db/version_set.cc

@ -2758,25 +2758,44 @@ void VersionStorageInfo::ComputeCompactionScore(
// Level-based involves L0->L0 compactions that can lead to oversized
// L0 files. Take into account size as well to avoid later giant
// compactions to the base level.
uint64_t l0_target_size = mutable_cf_options.max_bytes_for_level_base;
if (immutable_options.level_compaction_dynamic_level_bytes &&
level_multiplier_ != 0.0) {
// Prevent L0 to Lbase fanout from growing larger than
// `level_multiplier_`. This prevents us from getting stuck picking
// L0 forever even when it is hurting write-amp. That could happen
// in dynamic level compaction's write-burst mode where the base
// level's target size can grow to be enormous.
l0_target_size =
std::max(l0_target_size,
static_cast<uint64_t>(level_max_bytes_[base_level_] /
level_multiplier_));
}
score =
std::max(score, static_cast<double>(total_size) / l0_target_size);
if (immutable_options.level_compaction_dynamic_level_bytes &&
score > 1.0) {
// If score in L0 is always too high, L0->L1 will always be
// prioritized over L1->L2 compaction and L1 will accumulate to
// too large. But if L0 score isn't high enough, L0 will accumulate
// and data is not moved to L1 fast enough. With potential L0->L0
// compaction, number of L0 files aren't always an indication of
// L0 oversizing, and we also need to consider total size of L0.
if (immutable_options.level_compaction_dynamic_level_bytes) {
if (total_size >= mutable_cf_options.max_bytes_for_level_base) {
// When calculating estimated_compaction_needed_bytes, we assume
// L0 is qualified as pending compactions. We will need to make
// sure that it qualifies for compaction.
// It might be guafanteed by logic below anyway, but we are
// explicit here to make sure we don't stop writes with no
// compaction scheduled.
score = std::max(score, 1.01);
}
if (total_size > level_max_bytes_[base_level_]) {
// In this case, we compare L0 size with actual L1 size and make
// sure score is more than 1.0 (10.0 after scaled) if L0 is larger
// than L1. Since in this case L1 score is lower than 10.0, L0->L1
// is prioritized over L1->L2.
uint64_t base_level_size = 0;
for (auto f : files_[base_level_]) {
base_level_size += f->compensated_file_size;
}
score = std::max(score, static_cast<double>(total_size) /
static_cast<double>(std::max(
base_level_size,
level_max_bytes_[base_level_])));
}
if (score > 1.0) {
score *= kScoreScale;
}
} else {
score = std::max(score,
static_cast<double>(total_size) /
mutable_cf_options.max_bytes_for_level_base);
}
}
}
} else {

Loading…
Cancel
Save