Number of files explodes, compaction does not work..

[beviah]

tried numerous settings.. something does not work right..

there are thousands of tiny log and sst files ... not getting merged.

[Congyuwang]

which version are you using? And what platform are you using?

[beviah]

rocksdict 0.3.24
Python 3.12.3
Ubuntu 24.04.1 LTS

[Congyuwang]

That's kind of strange. Are you using too many column families maybe? Do you have a minimum code that can reproduce it?

[beviah]

I managed to get manual compaction working

def speedb_options():
    opt = Options()
    opt.create_if_missing(True)
    opt.create_missing_column_families(True)
    opt.set_max_open_files(-1)  # You don't have this set
    opt.set_max_background_jobs(4)
    opt.set_max_compaction_bytes(512 * 1024 * 1024)
    opt.set_max_subcompactions(4)
    opt.set_compaction_style(DBCompactionStyle.universal())
    opt.increase_parallelism(4)
    opt.set_use_direct_io_for_flush_and_compaction(True)
    opt.set_use_direct_reads(True)
    opt.set_writable_file_max_buffer_size(1024 * 1024)
    opt.set_write_buffer_size(64 * 1024 * 1024)
    opt.set_min_write_buffer_number(2)
    opt.set_max_write_buffer_number(6) 
    opt.set_min_write_buffer_number_to_merge(2)
    opt.set_target_file_size_base(64 * 1024 * 1024)
    opt.set_prefix_extractor(SliceTransform.create_max_len_prefix(8))
    opt.set_atomic_flush(True)
    return opt

i have 4 column families with above options. do not use defaults.

db = Rdict(
   shard_path, 
    speedb_options(), 
    column_families=column_families, 
    access_type=AccessType.read_write()
)

wb = WriteBatch()
wb.set_default_column_family(db.get_column_family_handle(x.cf))
for vid, content in vector_contents.items():
    wb[vid] = content
db.write(wb)

contents are just small jsons or lists of integers, depending on column family, vids are integers.

Will try to reproduce with separate minimal example.