mirror of
https://github.com/hanzoai/datastore.git
synced 2026-09-26 13:55:07 +00:00
324 lines
13 KiB
Python
324 lines
13 KiB
Python
#!/usr/bin/env python3
|
|
import json
|
|
import random
|
|
from pathlib import Path
|
|
|
|
from ci.jobs.ast_fuzzer_job import run_fuzz_job
|
|
from ci.jobs.ci_utils import is_extended_run
|
|
from ci.praktika.utils import Utils
|
|
|
|
|
|
def main():
|
|
temp_dir = Path(f"{Utils.cwd()}/ci/tmp/")
|
|
workspace_path = temp_dir / "workspace"
|
|
buzz_config_file = workspace_path / "fuzz.json"
|
|
|
|
workspace_path.mkdir(parents=True, exist_ok=True)
|
|
|
|
# Sometimes disallow SQL types to reduce number of combinations
|
|
disabled_types_str = ""
|
|
if random.randint(1, 2) == 1:
|
|
disabled_types = [
|
|
"bool",
|
|
"uint",
|
|
"int8",
|
|
"int16",
|
|
"int64",
|
|
"int128",
|
|
"bfloat16",
|
|
"float32",
|
|
"float64",
|
|
"date",
|
|
"date32",
|
|
"time",
|
|
"time64",
|
|
"datetime",
|
|
"datetime64",
|
|
"string",
|
|
"decimal",
|
|
"uuid",
|
|
"enum",
|
|
"dynamic",
|
|
"json",
|
|
"nullable",
|
|
"lcard",
|
|
"array",
|
|
"map",
|
|
"tuple",
|
|
"variant",
|
|
"nested",
|
|
"ipv4",
|
|
"ipv6",
|
|
"geo",
|
|
"fixedstring",
|
|
"qbit",
|
|
"aggregate",
|
|
"simpleaggregate",
|
|
]
|
|
random.shuffle(disabled_types)
|
|
disabled_types_str = ",".join(
|
|
[_ for _ in disabled_types[: random.randint(1, len(disabled_types))]]
|
|
)
|
|
|
|
# Sometimes disallow Table and Database engines to reduce number of combinations
|
|
disabled_engines_str = ""
|
|
if random.randint(1, 2) == 1:
|
|
disabled_engines = [
|
|
"replacingmergetree",
|
|
"coalescingmergetree",
|
|
"summingmergetree",
|
|
"aggregatingmergetree",
|
|
"collapsingmergetree",
|
|
"versionedcollapsingmergetree",
|
|
"file",
|
|
"null",
|
|
"set",
|
|
"join",
|
|
"memory",
|
|
"stripelog",
|
|
"log",
|
|
"tinylog",
|
|
"embeddedrocksdb",
|
|
"buffer",
|
|
"mysql",
|
|
"postgresql",
|
|
"sqlite",
|
|
"mongodb",
|
|
"redis",
|
|
"s3",
|
|
"s3queue",
|
|
"hudi",
|
|
"deltalakes3",
|
|
"deltalakeazure",
|
|
"deltalakelocal",
|
|
"icebergs3",
|
|
"icebergazure",
|
|
"iceberglocal",
|
|
"paimons3",
|
|
"paimonazure",
|
|
"paimonlocal",
|
|
"merge",
|
|
"distributed",
|
|
"dictionary",
|
|
"generaterandom",
|
|
"azureblobstorage",
|
|
"azurequeue",
|
|
"url",
|
|
"keepermap",
|
|
"externaldistributed",
|
|
"materializedpostgresql",
|
|
"replicated",
|
|
"shared",
|
|
"datalakecatalog",
|
|
"arrowflight",
|
|
"alias",
|
|
"kafka",
|
|
"backup",
|
|
]
|
|
random.shuffle(disabled_engines)
|
|
disabled_engines_str = ",".join(
|
|
[_ for _ in disabled_engines[: random.randint(1, len(disabled_engines))]]
|
|
)
|
|
|
|
# Generate configuration file
|
|
# No PARALLEL WITH with slow sanitizers
|
|
# If hardcoded inserts are allowed, reduce insert size, so logs don't grow as much
|
|
|
|
# TODO: Enable back after the issue with `assertHasValidVersionMetadata` will be fixed:
|
|
# https://play.clickhouse.com/play?user=play&run=1#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
|
|
allow_transactions = False
|
|
disallowed_settings = [
|
|
# Disable old analyzer always
|
|
"enable_analyzer",
|
|
# Don't always apply settings from the server
|
|
"apply_settings_from_server",
|
|
# Keep all debugging messages
|
|
"send_logs_level",
|
|
# Slow settings
|
|
"enable_producing_buckets_out_of_order_in_aggregation",
|
|
"grace_hash_join_initial_buckets",
|
|
"join_default_strictness",
|
|
"max_download_buffer_size",
|
|
"max_partitions_per_insert_block",
|
|
"max_table_size_to_drop",
|
|
"max_temporary_data_on_disk_size_for_query",
|
|
"max_temporary_data_on_disk_size_for_user",
|
|
"max_untracked_memory",
|
|
# Makes everything nullable, not handy
|
|
"data_type_default_nullable",
|
|
"restore_replace_external_dictionary_source_to_null",
|
|
"restore_replace_external_engines_to_null",
|
|
"restore_replace_external_table_functions_to_null",
|
|
# Not handy
|
|
"default_compression_codec",
|
|
"insert_shard_id",
|
|
"union_default_mode",
|
|
"except_default_mode",
|
|
"input_format_skip_unknown_fields",
|
|
"insert_quorum",
|
|
"temporary_files_buffer_size",
|
|
"trace_profile_events",
|
|
"unknown_packet_in_send_data",
|
|
]
|
|
if allow_transactions:
|
|
# Doesn't work well with transactions
|
|
disallowed_settings.append("apply_mutations_on_fly")
|
|
|
|
allow_hardcoded_inserts = random.choice([True, False])
|
|
min_nested_rows = random.randint(0, 5)
|
|
max_nested_rows = min_nested_rows + (5 if allow_hardcoded_inserts else 30)
|
|
min_insert_rows = random.randint(1, 100)
|
|
max_insert_rows = min_insert_rows + (10 if allow_hardcoded_inserts else 3000)
|
|
min_string_length = random.randint(0, 100)
|
|
max_string_length = min_string_length + (10 if allow_hardcoded_inserts else 300)
|
|
|
|
# Cap `max_depth` based on `max_nested_rows` so the worst-case nested-value
|
|
# product stays under ASan's allocation cap. BuzzHouse value generators
|
|
# for `Map` and `Array` recurse on their child types, multiplying the
|
|
# per-level row count at each nesting level. With `max_nested_rows` up to
|
|
# 105, depth 5 produces ~10^10 entries in a single value, which trips
|
|
# ASan's `allocation-size-too-big` guard.
|
|
#
|
|
# Worst-case nested-value product (`max_nested_rows ^ max_depth_high`):
|
|
# `max_nested_rows` <= 5 -> depth up to 5 (5^5 = 3125)
|
|
# `max_nested_rows` <= 20 -> depth up to 4 (20^4 = 160000)
|
|
# `max_nested_rows` > 20 -> depth up to 3 (~105^3 = ~1.2M)
|
|
# All bounded below ~2e6, well under ASan's ~1e9 allocation cap. Tiny
|
|
# `max_nested_rows` (0 or 1) keeps full depth-5 coverage for type-shape
|
|
# exploration; only the row-heavy branches lose the deepest levels.
|
|
if max_nested_rows <= 5:
|
|
max_depth_high = 5
|
|
elif max_nested_rows <= 20:
|
|
max_depth_high = 4
|
|
else:
|
|
max_depth_high = 3
|
|
|
|
buzz_config = {
|
|
"seed": random.randint(1, 18446744073709551615),
|
|
"max_depth": random.randint(2, max_depth_high),
|
|
"max_width": random.randint(2, 7),
|
|
"max_databases": random.randint(2, 5),
|
|
"max_tables": random.randint(3, 10),
|
|
"max_views": random.randint(0, 10),
|
|
"max_dictionaries": random.randint(0, 10),
|
|
"max_functions": random.randint(0, 8),
|
|
"max_policies": random.randint(0, 8),
|
|
"max_hypotheticals": random.randint(0, 0),
|
|
"max_columns": random.randint(1, 8),
|
|
"min_nested_rows": min_nested_rows,
|
|
"max_nested_rows": random.randint(min_nested_rows, max_nested_rows),
|
|
"min_insert_rows": min_insert_rows,
|
|
"max_insert_rows": random.randint(min_insert_rows, max_insert_rows),
|
|
"min_string_length": min_string_length,
|
|
"max_string_length": random.randint(min_string_length, max_string_length),
|
|
# Disable parallel queries, there are issues in the CI currently
|
|
# "max_parallel_queries": (
|
|
# 1
|
|
# if (
|
|
# any(x in Info().job_name for x in ["tsan", "asan", "msan"])
|
|
# or random.randint(1, 2) == 1
|
|
# )
|
|
# else random.randint(1, 5)
|
|
# ),
|
|
"max_parallel_queries": 1,
|
|
"max_number_alters": (1 if random.randint(1, 2) == 1 else random.randint(1, 4)),
|
|
"fuzz_floating_points": random.choice([True, False]),
|
|
"enable_fault_injection_settings": random.randint(1, 4) == 1,
|
|
"enable_force_settings": random.randint(1, 4) == 1,
|
|
"enable_time_settings": random.randint(1, 5) == 1,
|
|
# Don't compare for correctness yet, false positives maybe
|
|
"use_dump_table_oracle": (1 if random.randint(1, 3) == 1 else 0),
|
|
"test_with_fill": random.randint(1, 10) == 1,
|
|
"compare_success_results": False, # This can give false positives, so disable it
|
|
"allow_infinite_tables": random.randint(1, 10) == 1,
|
|
"allow_health_check": False, # I have to test this first
|
|
"allow_nasty_identifiers": random.randint(1, 8) == 1,
|
|
"enable_compatibility_settings": random.randint(1, 4) == 1,
|
|
"enable_memory_settings": random.randint(1, 4) == 1,
|
|
"enable_sync_settings": random.randint(1, 4) == 1,
|
|
"enable_backups": random.randint(1, 4) == 1,
|
|
"enable_renames": random.randint(1, 4) == 1,
|
|
"allow_hardcoded_inserts": allow_hardcoded_inserts,
|
|
"client_file_path": "/var/lib/clickhouse/user_files",
|
|
"server_file_path": "/var/lib/clickhouse/user_files",
|
|
"log_path": "/workspace/fuzzerout.sql",
|
|
"read_log": False,
|
|
"allow_memory_tables": random.choice([True, False]),
|
|
"allow_client_restarts": random.choice([True, False]),
|
|
# Sometimes use a small range of integer values
|
|
"random_limited_values": random.choice([True, False]),
|
|
"max_reconnection_attempts": 3,
|
|
"time_to_sleep_between_reconnects": 5000,
|
|
"keeper_map_path_prefix": "/keeper_map_tables",
|
|
# Oracles don't run check correctness on CI, so set deterministic probability low
|
|
"deterministic_prob": random.randint(0, 20),
|
|
"disabled_types": disabled_types_str,
|
|
"disabled_engines": disabled_engines_str,
|
|
# Make CI logs less verbose
|
|
"truncate_output": True,
|
|
# Don't always run transactions, makes many statements fail
|
|
"allow_transactions": allow_transactions,
|
|
# Run query oracles sometimes
|
|
"allow_query_oracles": random.randint(1, 4) == 1,
|
|
# Run for 30 minutes by default, or 60 minutes for feature/improvement/performance PRs
|
|
"time_to_run": 60 if is_extended_run() else 30,
|
|
"remote_servers": ["localhost:9000"],
|
|
"remote_secure_servers": ["localhost:9440"],
|
|
"http_servers": ["localhost:8123"],
|
|
"https_servers": ["localhost:8443"],
|
|
"disallowed_settings": disallowed_settings,
|
|
# MergeTree settings to set more often
|
|
"hot_table_settings": [
|
|
"add_minmax_index_for_block_number_column",
|
|
"add_minmax_index_for_block_offset_column",
|
|
"add_minmax_index_for_numeric_columns",
|
|
"add_minmax_index_for_string_columns",
|
|
"add_minmax_index_for_temporal_columns",
|
|
"allow_coalescing_columns_in_partition_or_order_key",
|
|
"allow_commit_order_projection",
|
|
"allow_experimental_text_index_phrase_search",
|
|
"allow_nullable_key",
|
|
"allow_part_offset_column_in_projections",
|
|
"allow_suspicious_indices",
|
|
"allow_tuple_element_aggregation",
|
|
"allow_vertical_merges_from_compact_to_wide_parts",
|
|
"auto_statistics_types",
|
|
"compress_marks",
|
|
"compress_primary_key",
|
|
"dynamic_serialization_version",
|
|
"enable_block_number_column",
|
|
"enable_block_offset_column",
|
|
"enable_index_granularity_compression",
|
|
"enable_vertical_merge_algorithm",
|
|
"index_granularity",
|
|
"index_granularity_bytes",
|
|
"map_buckets_strategy",
|
|
"map_serialization_version",
|
|
"merge_max_block_size",
|
|
"min_bytes_for_full_part_storage",
|
|
"min_bytes_for_wide_part",
|
|
"nullable_serialization_version",
|
|
"object_serialization_version",
|
|
"object_shared_data_buckets_for_compact_part",
|
|
"object_shared_data_buckets_for_wide_part",
|
|
"primary_key_lazy_load",
|
|
"ratio_of_defaults_for_sparse_serialization",
|
|
"serialization_info_version",
|
|
"share_nested_offsets",
|
|
"string_serialization_version",
|
|
"use_compact_variant_discriminators_serialization",
|
|
"use_const_adaptive_granularity",
|
|
"vertical_merge_algorithm_min_bytes_to_activate",
|
|
"vertical_merge_optimize_ttl_delete",
|
|
"write_marks_for_substreams_in_compact_parts",
|
|
],
|
|
}
|
|
with open(buzz_config_file, "w") as outfile:
|
|
outfile.write(json.dumps(buzz_config))
|
|
|
|
run_fuzz_job("BuzzHouse")
|
|
|
|
|
|
if __name__ == "__main__":
|
|
main()
|