Avoid full-range compactions with periodic filtered b.g. ones (#16697)

* Update rocksdb to v0.16.0

* Promote the infrequent and important log to info!

* Force background compaction by ttl without manual compaction

* Fix test

* Support no compaction mode in test_ledger_cleanup_compaction

* Fix comment

* Make compaction_interval customizable

* Avoid major compaction with periodic filtering...

* Adress lazy_static, special cfs and range check

* Clean up a bit and add comment

* Add comment

* More comments...

* Config code cleanup

* Add comment

* Use .conflicts_with()

* Nullify unneeded delete_range ops for special CFs

* Some clean ups

* Clarify the locking intention

* Ensure special CFs' consistency with PurgeType::CompactionFilter

* Fix comment

* Fix bad copy paste

* Fix various types...

* Don't use tuples

* Add a unit test for compaction_filter

* Fix typo...

* Remove flag and just use new behavior always

* Fix wrong condition negation...

* Doc. about no set_last_purged_slot in purge_slots

* Write a test and fix off-by-one bug....

* Apply suggestions from code review

Co-authored-by: Tyera Eulberg <teulberg@gmail.com>

* Follow up to github review suggestions

* Fix line-wrapping

* Fix conflict

Co-authored-by: Tyera Eulberg <teulberg@gmail.com>

core/src/ledger_cleanup_service.rs

@@ -207,11 +207,25 @@ impl LedgerCleanupService {
                     );
 
                     let mut purge_time = Measure::start("purge_slots");
+
                     blockstore.purge_slots(
                         purge_first_slot,
                         lowest_cleanup_slot,
-                        PurgeType::PrimaryIndex,
+                        PurgeType::CompactionFilter,
                     );
+                    // Update only after purge operation.
+                    // Safety: This value can be used by compaction_filters shared via Arc<AtomicU64>.
+                    // Compactions are async and run as a multi-threaded background job. However, this
+                    // shouldn't cause consistency issues for iterators and getters because we have
+                    // already expired all affected keys (older than or equal to lowest_cleanup_slot)
+                    // by the above `purge_slots`. According to the general RocksDB design where SST
+                    // files are immutable, even running iterators aren't affected; the database grabs
+                    // a snapshot of the live set of sst files at iterator's creation.
+                    // Also, we passed the PurgeType::CompactionFilter, meaning no delete_range for
+                    // transaction_status and address_signatures CFs. These are fine because they
+                    // don't require strong consistent view for their operation.
+                    blockstore.set_max_expired_slot(lowest_cleanup_slot);
+
                     purge_time.stop();
                     info!("{}", purge_time);
 

core/src/validator.rs

@@ -1636,9 +1636,11 @@ mod tests {
             }
             drop(blockstore);
 
+            // this purges and compacts all slots greater than or equal to 5
             backup_and_clear_blockstore(&blockstore_path, 5, 2);
 
             let blockstore = Blockstore::open(&blockstore_path).unwrap();
+            // assert that slots less than 5 aren't affected
             assert!(blockstore.meta(4).unwrap().unwrap().next_slots.is_empty());
             for i in 5..10 {
                 assert!(blockstore