// Licensed to the Apache Software Foundation (ASF) under one

// or more contributor license agreements.  See the NOTICE file

// distributed with this work for additional information

// regarding copyright ownership.  The ASF licenses this file

// to you under the Apache License, Version 2.0 (the

// "License"); you may not use this file except in compliance

// with the License.  You may obtain a copy of the License at

//

//   http://www.apache.org/licenses/LICENSE-2.0

//

// Unless required by applicable law or agreed to in writing,

// software distributed under the License is distributed on an

// "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY

// KIND, either express or implied.  See the License for the

// specific language governing permissions and limitations

// under the License.

#include <fmt/core.h>

#include <gflags/gflags.h>

#include <stdint.h>

#include <algorithm>

#include <cctype>

#include <cmath>

// IWYU pragma: no_include <bthread/errno.h>

#include <lz4/lz4hc.h>

#include <atomic>

#include <cerrno> // IWYU pragma: keep

#include <cmath>

#include <cstdlib>

#include <cstring>

#include <fstream> // IWYU pragma: keep

#include <functional>

#include <iostream>

#include <limits>

#include <map>

#include <memory>

#include <mutex>

#include <random>

#include <set>

#include <string>

#include <string_view>

#include <utility>

#include <vector>

#include "cloud/config.h"

#include "common/config.h"

#include "common/logging.h"

#include "common/status.h"

#include "io/fs/file_writer.h"

#include "io/fs/local_file_system.h"

#include "load/memtable/memtable_flush_executor.h"

#include "runtime/exec_env.h"

#include "runtime/workload_group/workload_group_manager.h"

#include "storage/storage_engine.h"

#include "util/cpu_info.h"

#include "util/string_util.h"

namespace doris::config {

#include "common/compile_check_avoid_begin.h"

// Dir of custom config file

DEFINE_String(custom_config_dir, "${DORIS_HOME}/conf");

// Dir of jdbc drivers

DEFINE_String(jdbc_drivers_dir, "${DORIS_HOME}/plugins/jdbc_drivers");

// cluster id

DEFINE_Int32(cluster_id, "-1");

// port on which BackendService is exported

DEFINE_Int32(be_port, "9060");

// port for brpc

DEFINE_Int32(brpc_port, "8060");

DEFINE_Int32(arrow_flight_sql_port, "8050");

// Validate Arrow input buffers in opted-in Arrow readers before converting them to Doris columns.

DEFINE_Bool(enable_arrow_input_validation, "true");

DEFINE_Int32(cdc_client_port, "9096");

DEFINE_String(cdc_client_java_opts, "");

// If the external client cannot directly access priority_networks, set public_host to be accessible

// to external client.

// There are usually two usage scenarios:

// 1. in production environment, it is often inconvenient to expose Doris BE nodes to the external network.

// However, a reverse proxy (such as Nginx) can be added to all Doris BE nodes, and the external client will be

// randomly routed to a Doris BE node when connecting to Nginx. set public_host to the host of Nginx.

// 2. if priority_networks is an internal network IP, and BE node has its own independent external IP,

// but Doris currently does not support modifying priority_networks, setting public_host to the real external IP.

DEFINE_mString(public_host, "");

// If the BE node is connected to the external network through a reverse proxy like Nginx

// and need to use Arrow Flight SQL, should add a server in Nginx to reverse proxy

// `Nginx:arrow_flight_sql_proxy_port` to `BE_priority_networks:arrow_flight_sql_port`. For example:

// upstream arrowflight {

//    server 10.16.10.8:8069;

//    server 10.16.10.8:8068;

//}

// server {

//    listen 8167 http2;

//    listen [::]:8167 http2;

//    server_name doris.arrowflight.com;

// }

DEFINE_Int32(arrow_flight_sql_proxy_port, "-1");

// the number of bthreads for brpc, the default value is set to -1,

// which means the number of bthreads is #cpu-cores

DEFINE_Int32(brpc_num_threads, "256");

// the time of brpc server keep idle connection, setting this value too small may cause rpc between backends to fail,

// the default value is set to -1, which means never close idle connection.

DEFINE_Int32(brpc_idle_timeout_sec, "-1");

// Maximum number of multi-dimensional bvar samples exported by /brpc_metrics.

// brpc 1.4 defaults this flag to 0, which disables all MultiDimension metrics.

DEFINE_String(bvar_max_dump_multi_dimension_metric_num, "5000");

// Declare a selection strategy for those servers have many ips.

// Note that there should at most one ip match this list.

// this is a list in semicolon-delimited format, in CIDR notation, e.g. 10.10.10.0/24

// If no ip match this rule, will choose one randomly.

DEFINE_String(priority_networks, "");

// memory mode

// performance or compact

DEFINE_String(memory_mode, "moderate");

DEFINE_mBool(enable_use_cgroup_memory_info, "true");

// process memory limit specified as number of bytes

// ('<int>[bB]?'), megabytes ('<float>[mM]'), gigabytes ('<float>[gG]'),

// or percentage of the physical memory ('<int>%').

// defaults to bytes if no unit is given"

// must larger than 0. and if larger than physical memory size,

// it will be set to physical memory size.

DEFINE_String(mem_limit, "90%");

// Soft memory limit as a fraction of hard memory limit.

DEFINE_Double(soft_mem_limit_frac, "0.9");

// Cache capacity reduce mem limit as a fraction of soft mem limit.

DEFINE_mDouble(cache_capacity_reduce_mem_limit_frac, "0.7");

// Schema change memory limit as a fraction of soft memory limit.

DEFINE_Double(schema_change_mem_limit_frac, "0.6");

// Many modern allocators (for example, tcmalloc) do not do a mremap for

// realloc, even in case of large enough chunks of memory. Although this allows

// you to increase performance and reduce memory consumption during realloc.

// To fix this, we do mremap manually if the chunk of memory is large enough.

//

// The threshold (128 MB, 128 * (1ULL << 20)) is chosen quite large, since changing the address

// space is very slow, especially in the case of a large number of threads. We

// expect that the set of operations mmap/something to do/mremap can only be

// performed about 1000 times per second.

//

// P.S. This is also required, because tcmalloc can not allocate a chunk of

// memory greater than 16 GB.

DEFINE_mInt64(mmap_threshold, "134217728"); // bytes

// When hash table capacity is greater than 2^double_grow_degree(default 2G), grow when 75% of the capacity is satisfied.

// Increase can reduce the number of hash table resize, but may waste more memory.

DEFINE_mInt32(hash_table_double_grow_degree, "31");

DEFINE_mInt32(max_fill_rate, "2");

DEFINE_mInt32(double_resize_threshold, "23");

// The maximum low water mark of the system `/proc/meminfo/MemAvailable`, Unit byte, default -1.

// if it is -1, then low water mark = min(MemTotal - MemLimit, MemTotal * 5%), which is 3.2G on a 64G machine.

// Turn up max. more memory buffers will be reserved for Memory GC.

// Turn down max. will use as much memory as possible.

// note that: `max_` prefix should be removed, but keep it for compatibility.

DEFINE_Int64(max_sys_mem_available_low_water_mark_bytes, "-1");

DEFINE_Int64(memtable_limiter_reserved_memory_bytes, "838860800");

// The size of the memory that gc wants to release each time, as a percentage of the mem limit.

DEFINE_mString(process_minor_gc_size, "5%");

DEFINE_mString(process_full_gc_size, "10%");

// gc will release cache, cancel task, and task will wait for gc to release memory,

// default gc strategy is conservative, if you want to exclude the interference of gc, let it be true

DEFINE_mBool(disable_memory_gc, "false");

// for the query being canceled,

// if (current time - cancel start time) < revoke_memory_max_tolerance_ms, the query memory is counted in `freed_memory`,

// and the query memory is expected to be released soon.

// if > revoke_memory_max_tolerance_ms, the query memory will not be counted in `freed_memory`,

// and the query may be blocked during the cancel process. skip this query and continue to cancel other queries.

DEFINE_mInt64(revoke_memory_max_tolerance_ms, "3000");

DEFINE_mBool(enable_stacktrace, "true");

DEFINE_mInt64(stacktrace_in_alloc_large_memory_bytes, "2147483647"); // 2GB -1

DEFINE_mInt64(crash_in_alloc_large_memory_bytes, "-1");

// The actual meaning of this parameter is `debug_memory`.

// 1. crash in memory tracker inaccurate, if memory tracker value is inaccurate, BE will crash.

//    usually used in test environments, default value is false.

// 2. print more memory logs.

DEFINE_mBool(crash_in_memory_tracker_inaccurate, "false");

// default is true. if any memory tracking in Orphan mem tracker will report error.

// !! not modify the default value of this conf!! otherwise memory errors cannot be detected in time.

// allocator free memory not need to check, because when the thread memory tracker label is Orphan,

// use the tracker saved in Allocator.

DEFINE_mBool(enable_memory_orphan_check, "true");

// The maximum time a thread waits for full GC. Currently only query will wait for full gc.

DEFINE_mInt32(thread_wait_gc_max_milliseconds, "1000");

DEFINE_mInt64(pre_serialize_keys_limit_bytes, "16777216");

// the port heartbeat service used

DEFINE_Int32(heartbeat_service_port, "9050");

// the count of heart beat service

DEFINE_Int32(heartbeat_service_thread_count, "1");

// the count of thread to create table

DEFINE_Int32(create_tablet_worker_count, "3");

// the count of thread to drop table

DEFINE_Int32(drop_tablet_worker_count, "3");

// the count of thread to batch load

DEFINE_Int32(push_worker_count_normal_priority, "3");

// the count of thread to high priority batch load

DEFINE_Int32(push_worker_count_high_priority, "3");

// the count of thread to publish version

DEFINE_Int32(publish_version_worker_count, "8");

// the count of tablet thread to publish version

DEFINE_Int32(tablet_publish_txn_max_thread, "32");

// the timeout of EnginPublishVersionTask

DEFINE_Int32(publish_version_task_timeout_s, "8");

// the count of thread to calc delete bitmap

DEFINE_Int32(calc_delete_bitmap_max_thread, "32");

// the num of threads to calc delete bitmap when building rowset, 0 = auto

DEFINE_Int32(calc_delete_bitmap_for_load_max_thread, "0");

// the count of thread to calc delete bitmap worker, only used for cloud

DEFINE_Int32(calc_delete_bitmap_worker_count, "8");

// the count of thread to calc tablet delete bitmap task, only used for cloud

DEFINE_Int32(calc_tablet_delete_bitmap_task_max_thread, "32");

// the count of thread to clear transaction task

DEFINE_Int32(clear_transaction_task_worker_count, "1");

// the count of thread to delete

DEFINE_Int32(delete_worker_count, "3");

// the count of thread to alter table

DEFINE_Int32(alter_tablet_worker_count, "3");

// the count of thread to alter index

DEFINE_Int32(alter_index_worker_count, "3");

// the count of thread to clone

DEFINE_Int32(clone_worker_count, "3");

// the count of thread to clone

DEFINE_Int32(storage_medium_migrate_count, "1");

// the count of thread to check consistency

DEFINE_Int32(check_consistency_worker_count, "1");

// the count of thread to upload

DEFINE_Int32(upload_worker_count, "1");

// the count of thread to download

DEFINE_Int32(download_worker_count, "1");

DEFINE_Int32(num_query_ctx_map_partitions, "128");

// the count of thread to make snapshot

DEFINE_Int32(make_snapshot_worker_count, "5");

// the count of thread to release snapshot

DEFINE_Int32(release_snapshot_worker_count, "5");

// the count of thread to make committed rowsets visible in cloud mode

DEFINE_Int32(cloud_make_committed_rs_visible_worker_count, "16");

// report random wait a little time to avoid FE receiving multiple be reports at the same time.

// do not set it to false for production environment

DEFINE_mBool(report_random_wait, "true");

// the interval time(seconds) for agent report tasks signature to FE

DEFINE_mInt32(report_task_interval_seconds, "10");

// the interval time(seconds) for agent report disk state to FE

DEFINE_mInt32(report_disk_state_interval_seconds, "30");

// the interval time(seconds) for agent report olap table to FE

DEFINE_mInt32(report_tablet_interval_seconds, "60");

// the max download speed(KB/s)

DEFINE_mInt32(max_download_speed_kbps, "50000");

// download low speed limit(KB/s)

DEFINE_mInt32(download_low_speed_limit_kbps, "50");

// download low speed time(seconds)

DEFINE_mInt32(download_low_speed_time, "300");

// whether to download small files in batch

DEFINE_mBool(enable_batch_download, "true");

// whether to enable stream load forward endpoint for cloud group commit

DEFINE_mBool(enable_group_commit_streamload_be_forward, "false");

// whether to check md5sum when download

DEFINE_mBool(enable_download_md5sum_check, "false");

// download binlog meta timeout, default 30s

DEFINE_mInt32(download_binlog_meta_timeout_ms, "30000");

// the interval time(seconds) for agent report index policy to FE

DEFINE_mInt32(report_index_policy_interval_seconds, "10");

// DNS cache: throttle "use cached ip" warning to once per N failures per host.

DEFINE_mInt32(dns_cache_log_every_n_failures, "10");

// DNS cache: evict a hostname after this many consecutive resolution failures.

DEFINE_mInt32(dns_cache_max_consecutive_failures, "30");

// DNS cache: after eviction, block re-resolve attempts for this many seconds.

// Set <= 0 to disable the negative cache (legacy behavior).

DEFINE_mInt32(dns_cache_negative_ttl_seconds, "60");

DEFINE_String(sys_log_dir, "");

DEFINE_String(user_function_dir, "${DORIS_HOME}/lib/udf");

// INFO, WARNING, ERROR, FATAL

DEFINE_mString(sys_log_level, "INFO");

// TIME-DAY, TIME-HOUR, SIZE-MB-nnn

DEFINE_String(sys_log_roll_mode, "SIZE-MB-1024");

// log roll num

DEFINE_Int32(sys_log_roll_num, "10");

// verbose log

DEFINE_Strings(sys_log_verbose_modules, "");

// verbose log level

DEFINE_Int32(sys_log_verbose_level, "10");

// verbose log FLAGS_v

DEFINE_Int32(sys_log_verbose_flags_v, "-1");

// log buffer level

DEFINE_String(log_buffer_level, "");

// log enable custom date time format

DEFINE_Bool(sys_log_enable_custom_date_time_format, "false");

// log custom date time format (https://en.cppreference.com/w/cpp/io/manip/put_time)

DEFINE_String(sys_log_custom_date_time_format, "%Y-%m-%d %H:%M:%S");

// log custom date time milliseconds format (fmt::format)

DEFINE_String(sys_log_custom_date_time_ms_format, ",{:03d}");

// number of threads available to serve backend execution requests

DEFINE_Int32(be_service_threads, "64");

// The pipeline task has a high concurrency, therefore reducing its report frequency

DEFINE_mInt32(pipeline_status_report_interval, "10");

DEFINE_mInt32(pipeline_task_exec_time_slice, "100");

// task executor min concurrency per task

DEFINE_Int32(task_executor_min_concurrency_per_task, "1");

// task executor max concurrency per task

DEFINE_Int32(task_executor_max_concurrency_per_task, "-1");

// task task executor inital split max concurrency per task, later concurrency may be adjusted dynamically

DEFINE_Int32(task_executor_initial_max_concurrency_per_task, "-1");

// Enable task executor in internal table scan.

DEFINE_Bool(enable_task_executor_in_internal_table, "false");

// Enable task executor in external table scan.

DEFINE_Bool(enable_task_executor_in_external_table, "false");

// number of scanner thread pool size for olap table

// and the min thread num of remote scanner thread pool

DEFINE_Int32(doris_scanner_thread_pool_thread_num, "-1");

DEFINE_Int32(doris_scanner_min_thread_pool_thread_num, "8");

DEFINE_Int32(remote_split_source_batch_size, "1000");

DEFINE_Int32(doris_max_remote_scanner_thread_pool_thread_num, "-1");

// number of olap scanner thread pool queue size

DEFINE_Int32(doris_scanner_thread_pool_queue_size, "102400");

// default thrift client connect timeout(in seconds)

DEFINE_mInt32(thrift_connect_timeout_seconds, "3");

// default thrift client retry interval (in milliseconds)

DEFINE_mInt64(thrift_client_retry_interval_ms, "1000");

// max message size of thrift request

// default: 100 * 1024 * 1024

DEFINE_mInt32(thrift_max_message_size, "104857600");

// max bytes number for single scan range, used in segmentv2

DEFINE_mInt32(doris_scan_range_max_mb, "1024");

// single read execute fragment row number

DEFINE_mInt32(doris_scanner_row_num, "16384");

// single read execute fragment row bytes

DEFINE_mInt32(doris_scanner_row_bytes, "10485760");

DEFINE_mInt32(doris_scanner_dynamic_interval_ms, "100");

// (Advanced) Maximum size of per-query receive-side buffer

DEFINE_mInt32(exchg_node_buffer_size_bytes, "20485760");

DEFINE_mInt32(exchg_buffer_queue_capacity_factor, "64");

// memory_limitation_per_thread_for_schema_change_bytes unit bytes

DEFINE_mInt64(memory_limitation_per_thread_for_schema_change_bytes, "2147483648");

DEFINE_mInt32(cache_prune_interval_sec, "10");

DEFINE_mInt32(cache_periodic_prune_stale_sweep_sec, "60");

// the clean interval of tablet lookup cache

DEFINE_mInt32(tablet_lookup_cache_stale_sweep_time_sec, "30");

DEFINE_mInt32(point_query_row_cache_stale_sweep_time_sec, "300");

DEFINE_mInt32(disk_stat_monitor_interval, "5");

DEFINE_mInt32(unused_rowset_monitor_interval, "30");

// Legacy name retained for compatibility; controls GLOBAL_ROWID_COL file-map GC.

DEFINE_mInt32(quering_rowsets_evict_interval, "30");

DEFINE_String(storage_root_path, "${DORIS_HOME}/storage");

DEFINE_mString(broken_storage_path, "");

DEFINE_Int32(min_active_scan_threads, "-1");

DEFINE_Int32(min_active_file_scan_threads, "-1");

// Config is used to check incompatible old format hdr_ format

// whether doris uses strict way. When config is true, process will log fatal

// and exit. When config is false, process will only log warning.

DEFINE_Bool(storage_strict_check_incompatible_old_format, "true");

// BE process will exit if the percentage of error disk reach this value.

DEFINE_mInt32(max_percentage_of_error_disk, "100");

DEFINE_mInt32(default_num_rows_per_column_file_block, "1024");

// pending data policy

DEFINE_mInt32(pending_data_expire_time_sec, "1800");

// inc_rowset snapshot rs sweep time interval

DEFINE_mInt32(tablet_rowset_stale_sweep_time_sec, "600");

// tablet stale rowset sweep by threshold size

DEFINE_Bool(tablet_rowset_stale_sweep_by_size, "false");

DEFINE_mInt32(tablet_rowset_stale_sweep_threshold_size, "100");

// garbage sweep policy

DEFINE_Int32(max_garbage_sweep_interval, "3600");

DEFINE_Int32(min_garbage_sweep_interval, "180");

DEFINE_mInt32(garbage_sweep_batch_size, "100");

DEFINE_mInt32(snapshot_expire_time_sec, "172800");

// It is only a recommended value. When the disk space is insufficient,

// the file storage period under trash dose not have to comply with this parameter.

DEFINE_mInt32(trash_file_expire_time_sec, "0");

// minimum file descriptor number

// modify them upon necessity

DEFINE_Int32(min_file_descriptor_number, "60000");

DEFINE_mBool(disable_segment_cache, "false");

// Enable checking segment rows consistency between rowset meta and segment footer

DEFINE_mBool(enable_segment_rows_consistency_check, "false");

DEFINE_mBool(enable_segment_rows_check_core, "false");

// ATTENTION: For test only. In test environment, there are no historical data,

// so all rowset meta should have segment rows info.

DEFINE_mBool(fail_when_segment_rows_not_in_rowset_meta, "false");

DEFINE_String(row_cache_mem_limit, "20%");

// Cache for storage page size

DEFINE_String(storage_page_cache_limit, "20%");

// Shard size for page cache, the value must be power of two.

// It's recommended to set it to a value close to the number of BE cores in order to reduce lock contentions.

DEFINE_Int32(storage_page_cache_shard_size, "256");

DEFINE_mInt32(file_cache_mem_storage_shard_num, "1024");

// Percentage for index page cache

// all storage page cache will be divided into data_page_cache and index_page_cache

DEFINE_Int32(index_page_cache_percentage, "10");

// whether to disable page cache feature in storage

DEFINE_mBool(disable_storage_page_cache, "false");

// whether to disable row cache feature in storage

DEFINE_mBool(disable_storage_row_cache, "true");

// Parquet page cache: threshold ratio for caching decompressed vs compressed pages

// If uncompressed_size / compressed_size <= threshold, cache decompressed;

// otherwise cache compressed if enable_parquet_cache_compressed_pages = true

DEFINE_Double(parquet_page_cache_decompress_threshold, "1.5");

// Parquet page cache: whether to enable caching compressed pages (when ratio exceeds threshold)

DEFINE_Bool(enable_parquet_cache_compressed_pages, "false");

// whether to disable pk page cache feature in storage

DEFINE_Bool(disable_pk_storage_page_cache, "false");

// Cache for mow primary key storage page size

DEFINE_String(pk_storage_page_cache_limit, "10%");

// data page size for primary key index

DEFINE_Int32(primary_key_data_page_size, "32768");

DEFINE_mInt32(data_page_cache_stale_sweep_time_sec, "300");

DEFINE_mInt32(index_page_cache_stale_sweep_time_sec, "600");

DEFINE_mInt32(pk_index_page_cache_stale_sweep_time_sec, "600");

DEFINE_mBool(enable_low_cardinality_optimize, "true");

DEFINE_Bool(enable_low_cardinality_cache_code, "true");

DEFINE_mBool(enable_adaptive_batch_size, "true");

// be policy

// whether check compaction checksum

DEFINE_mBool(enable_compaction_checksum, "false");

// whether disable automatic compaction task

DEFINE_mBool(disable_auto_compaction, "false");

// whether enable vertical compaction

DEFINE_mBool(enable_vertical_compaction, "true");

// whether enable ordered data compaction

DEFINE_mBool(enable_ordered_data_compaction, "true");

// In vertical compaction, column number for every group

DEFINE_mInt32(vertical_compaction_num_columns_per_group, "5");

// In vertical compaction, max memory usage for row_source_buffer

DEFINE_Int32(vertical_compaction_max_row_source_memory_mb, "1024");

// In vertical compaction, max dest segment file size

DEFINE_mInt64(vertical_compaction_max_segment_size, "1073741824");

// Density threshold for sparse column compaction optimization

// density = (total_cells - null_cells) / total_cells, smaller means more sparse

// When density <= threshold, enable sparse optimization

// 0 = disable optimization, 1 = always enable

// Default 0.05 means enable sparse optimization when desity <= 5%

DEFINE_mDouble(sparse_column_compaction_threshold_percent, "0.05");

// Enable RLE batch Put optimization for compaction

DEFINE_mBool(enable_rle_batch_put_optimization, "true");

// Enable PDEP-based bit unpacking. Disable it on CPUs where PDEP is microcoded and slower than

// the scalar implementation, such as AMD Zen+ and Zen 2.

DEFINE_Bool(enable_bmi2_optimizations, "true");

// In ordered data compaction, min segment size for input rowset

DEFINE_mInt32(ordered_data_compaction_min_segment_size, "10485760");

// This config can be set to limit thread number in compaction thread pool.

DEFINE_mInt32(max_base_compaction_threads, "4");

DEFINE_mInt32(max_cumu_compaction_threads, "-1");

// Changelog compaction

// ATTN(changelog-rename): formerly "binlog_compaction_wait_timesec_after_visible"

DEFINE_mInt64(changelog_compaction_wait_timesec_after_visible, "600");

// ATTN(changelog-rename): formerly "binlog_compaction_goal_size_mbytes"

DEFINE_mInt64(changelog_compaction_goal_size_mbytes, "128");

// ATTN(changelog-rename): formerly "binlog_compaction_task_num_per_disk"

DEFINE_mInt32(changelog_compaction_task_num_per_disk, "4");

// ATTN(changelog-rename): formerly "binlog_compaction_file_count_threshold"

DEFINE_mInt32(changelog_compaction_file_count_threshold, "100");

// ATTN(changelog-rename): formerly "binlog_level_compaction_max_deltas"

DEFINE_mInt32(changelog_level_compaction_max_deltas, "2000");

// ATTN(changelog-rename): formerly "binlog_compaction_time_threshold_seconds"

DEFINE_mInt64(changelog_compaction_time_threshold_seconds, "3600");

// ATTN(changelog-rename): formerly "max_binlog_compaction_threads"

DEFINE_mInt32(max_changelog_compaction_threads, "-1");

DEFINE_Bool(enable_base_compaction_idle_sched, "true");

DEFINE_mInt64(base_compaction_min_rowset_num, "5");

DEFINE_mInt64(base_compaction_max_compaction_score, "20");

DEFINE_mInt64(mow_base_compaction_max_compaction_score, "200");

DEFINE_mDouble(base_compaction_min_data_ratio, "0.3");

DEFINE_mInt64(base_compaction_dup_key_max_file_size_mbytes, "1024");

DEFINE_Bool(enable_skip_tablet_compaction, "true");

DEFINE_mInt32(skip_tablet_compaction_second, "10");

// output rowset of cumulative compaction total disk size exceed this config size,

// this rowset will be given to base compaction, unit is m byte.

DEFINE_mInt64(compaction_promotion_size_mbytes, "1024");

// output rowset of cumulative compaction total disk size exceed this config ratio of

// base rowset's total disk size, this rowset will be given to base compaction. The value must be between

// 0 and 1.

DEFINE_mDouble(compaction_promotion_ratio, "0.05");

// the smallest size of rowset promotion. When the rowset is less than this config, this

// rowset will be not given to base compaction. The unit is m byte.

DEFINE_mInt64(compaction_promotion_min_size_mbytes, "128");

// When output rowset of cumulative compaction total version count (end_version - start_version)

// exceed this config count, the rowset will be moved to base compaction.

// NOTE: this config only works for unique key merge-on-write tables. The default maximum value

// disables version-count-based promotion while retaining the option to enable it when needed.

DEFINE_mInt64(compaction_promotion_version_count, "9223372036854775807");

// The lower bound size to do cumulative compaction. When total disk size of candidate rowsets is less than

// this size, size_based policy may not do to cumulative compaction. The unit is m byte.

DEFINE_mInt64(compaction_min_size_mbytes, "64");

// cumulative compaction policy: min and max delta file's number

DEFINE_mInt64(cumulative_compaction_min_deltas, "5");

DEFINE_mInt64(cumulative_compaction_max_deltas, "1000");

DEFINE_mInt32(cumulative_compaction_max_deltas_factor, "10");

// The upper limit of "permits" held by all compaction tasks. This config can be set to limit memory consumption for compaction.

DEFINE_mInt64(total_permits_for_compaction_score, "1000000");

// sleep interval in ms after generated compaction tasks

DEFINE_mInt32(generate_compaction_tasks_interval_ms, "100");

// Compaction task number per disk.

// Must be greater than 2, because Base compaction and Cumulative compaction have at least one thread each.

DEFINE_mInt32(compaction_task_num_per_disk, "4");

// compaction thread num for fast disk(typically .SSD), must be greater than 2.

DEFINE_mInt32(compaction_task_num_per_fast_disk, "8");

DEFINE_Validator(compaction_task_num_per_disk,

                 [](const int config) -> bool { return config >= 2; });

DEFINE_Validator(compaction_task_num_per_fast_disk,

                 [](const int config) -> bool { return config >= 2; });

DEFINE_Validator(low_priority_compaction_task_num_per_disk,

                 [](const int config) -> bool { return config >= 2; });

// How many rounds of cumulative compaction for each round of base compaction when compaction tasks generation.

DEFINE_mInt32(cumulative_compaction_rounds_for_each_base_compaction_round, "9");

// Minimum number of threads required in the thread pool to activate the large cumu compaction delay strategy.

// The delay strategy is only applied when the thread pool has at least this many threads.

// Default -1 means disable.

DEFINE_mInt32(large_cumu_compaction_task_min_thread_num, "-1");

// Maximum size threshold (in bytes) for input rowsets. Compaction tasks with input size

// exceeding this threshold will be delayed when thread pool is near capacity. Default 512MB.

DEFINE_mInt32(large_cumu_compaction_task_bytes_threshold, "536870912");

// Maximum row count threshold for compaction input. Compaction tasks with row count

// exceeding this threshold will be delayed when thread pool is near capacity. Default 1 million.

DEFINE_mInt32(large_cumu_compaction_task_row_num_threshold, "1000000");

// Not compact the invisible versions, but with some limitations:

// if not timeout, keep no more than compaction_keep_invisible_version_max_count versions;

// if timeout, keep no more than compaction_keep_invisible_version_min_count versions.

DEFINE_mInt32(compaction_keep_invisible_version_timeout_sec, "1800");

DEFINE_mInt32(compaction_keep_invisible_version_min_count, "50");

DEFINE_mInt32(compaction_keep_invisible_version_max_count, "500");

// Threshold to logging compaction trace, in seconds.

DEFINE_mInt32(base_compaction_trace_threshold, "60");

DEFINE_mInt32(cumulative_compaction_trace_threshold, "10");

DEFINE_mBool(disable_compaction_trace_log, "true");

DEFINE_mBool(enable_compaction_task_tracker, "true");

DEFINE_mInt32(compaction_task_tracker_max_records, "10000");

// Interval to picking rowset to compact, in seconds

DEFINE_mInt64(pick_rowset_to_compact_interval_sec, "86400");

// Compaction priority schedule

DEFINE_mBool(enable_compaction_priority_scheduling, "true");

DEFINE_mInt32(low_priority_compaction_task_num_per_disk, "2");

DEFINE_mInt32(low_priority_compaction_score_threshold, "200");

// Thread count to do tablet meta checkpoint, -1 means use the data directories count.

DEFINE_Int32(max_meta_checkpoint_threads, "-1");

// Threshold to logging agent task trace, in seconds.

DEFINE_mInt32(agent_task_trace_threshold_sec, "2");

// This config can be set to limit thread number in tablet migration thread pool.

DEFINE_Int32(min_tablet_migration_threads, "1");

DEFINE_Int32(max_tablet_migration_threads, "1");

DEFINE_mInt32(finished_migration_tasks_size, "10000");

// If size less than this, the remaining rowsets will be force to complete

DEFINE_mInt32(migration_remaining_size_threshold_mb, "10");

// If the task runs longer than this time, the task will be terminated, in seconds.

// timeout = std::max(migration_task_timeout_secs,  tablet size / 1MB/s)

DEFINE_mInt32(migration_task_timeout_secs, "300");

// timeout for try_lock migration lock

DEFINE_Int64(migration_lock_timeout_ms, "1000");

// Port to start debug webserver on

DEFINE_Int32(webserver_port, "8040");

// TLS module enable flag

DEFINE_Bool(enable_tls, "false");

// Path of TLS certificate

DEFINE_String(tls_certificate_path, "");

// Path of TLS private key

DEFINE_String(tls_private_key_path, "");

// Password for encrypted TLS private key

DEFINE_String(tls_private_key_password, "");

// TLS peer verification mode

DEFINE_String(tls_verify_mode, "verify_peer");

// Path of TLS CA certificate

DEFINE_String(tls_ca_certificate_path, "");

// TLS certificate reload interval, in seconds

DEFINE_Int32(tls_cert_refresh_interval_seconds, "3600");

// Comma-separated excluded server protocols: brpc,thrift,http,arrowflight

DEFINE_String(tls_excluded_protocols, "");

// Required peer certificate DNS SAN allowlist for private protocols, syntax: brpc=a.com;thrift=b.com.

// Empty means allow all peers. Once configured, the list acts as an allowlist and only peers whose

// DNS SAN matches at least one configured entry for that protocol are allowed.

DEFINE_String(tls_peer_cert_required_san_dns, "");

// Https enable flag

DEFINE_Bool(enable_https, "false");

// Path of certificate

DEFINE_String(ssl_certificate_path, "");

// Path of private key

DEFINE_String(ssl_private_key_path, "");

// Whether to check authorization

DEFINE_Bool(enable_all_http_auth, "false");

// Number of webserver workers

DEFINE_Int32(webserver_num_workers, "128");

// Async replies: stream load only now

// reply wait timeout only happens if:

// 1. Stream load fragment execution times out

//    HTTP request freed → stream load canceled

// 2. Client disconnects

DEFINE_mInt32(async_reply_timeout_s, "60");

DEFINE_Validator(async_reply_timeout_s, [](const int config) -> bool { return config >= 3; });

// Used for mini Load. mini load data file will be removed after this time.

DEFINE_Int64(load_data_reserve_hours, "4");

// log error log will be removed after this time

DEFINE_mInt64(load_error_log_reserve_hours, "48");

// error log size limit, default 200MB

DEFINE_mInt64(load_error_log_limit_bytes, "209715200");

// Dedicated load cancellation workers. Requires a restart.

DEFINE_Int32(brpc_load_light_work_pool_threads, "32");

DEFINE_Validator(brpc_load_light_work_pool_threads,

                 [](const int config) -> bool { return config > 0; });

// Queue capacity: -1 selects a CPU-scaled default. Requires a restart.

DEFINE_Int32(brpc_load_light_work_pool_max_queue_size, "-1");

DEFINE_Validator(brpc_load_light_work_pool_max_queue_size,

                 [](const int config) -> bool { return config == -1 || config > 0; });

DEFINE_Int32(brpc_heavy_work_pool_threads, "-1");

DEFINE_Int32(brpc_peer_fetch_pool_threads, "-1");

DEFINE_Int32(brpc_light_work_pool_threads, "-1");

DEFINE_Int32(brpc_heavy_work_pool_max_queue_size, "-1");

DEFINE_Int32(brpc_peer_fetch_pool_max_queue_size, "-1");

DEFINE_Int32(brpc_light_work_pool_max_queue_size, "-1");

DEFINE_mBool(enable_bthread_transmit_block, "true");

DEFINE_Int32(brpc_arrow_flight_work_pool_threads, "-1");

DEFINE_Int32(brpc_arrow_flight_work_pool_max_queue_size, "-1");

//Enable brpc builtin services, see:

//https://brpc.apache.org/docs/server/basics/#disable-built-in-services-completely

DEFINE_Bool(enable_brpc_builtin_services, "true");

// Enable brpc connection check

DEFINE_Bool(enable_brpc_connection_check, "false");

DEFINE_mInt64(brpc_connection_check_timeout_ms, "10000");

// The maximum amount of data that can be processed by a stream load

DEFINE_mInt64(streaming_load_max_mb, "102400");

// Some data formats, such as JSON, cannot be streamed.

// Therefore, it is necessary to limit the maximum number of

// such data when using stream load to prevent excessive memory consumption.

DEFINE_mInt64(streaming_load_json_max_mb, "100");

// the alive time of a TabletsChannel.

// If the channel does not receive any data till this time,

// the channel will be removed.

DEFINE_mInt32(streaming_load_rpc_max_alive_time_sec, "1200");

// the timeout of a rpc to open the tablet writer in remote BE.

// short operation time, can set a short timeout

DEFINE_Int32(tablet_writer_open_rpc_timeout_sec, "60");

// You can ignore brpc error '[E1011]The server is overcrowded' when writing data.

DEFINE_mBool(tablet_writer_ignore_eovercrowded, "true");

// Whether to enable stream load record function, the default is false.

// False: disable stream load record

DEFINE_mBool(enable_stream_load_record, "false");

// Whether to enable stream load record to audit log table, the default is true.

DEFINE_mBool(enable_stream_load_record_to_audit_log_table, "false");

// the maximum bytes of a batch of stream load records to audit log table

DEFINE_mInt64(stream_load_record_batch_bytes, "104857600"); // 100MB

// the interval to send a batch of stream load records to audit log table

DEFINE_mInt64(stream_load_record_batch_interval_secs, "120"); // 2 minutes

// batch size of stream load record reported to FE

DEFINE_mInt32(stream_load_record_batch_size, "50");

// expire time of stream load record in rocksdb.

DEFINE_Int32(stream_load_record_expire_time_secs, "28800");

// time interval to clean expired stream load records

DEFINE_mInt64(clean_stream_load_record_interval_secs, "1800");

// enable stream load commit txn on BE directly, bypassing FE. Only for cloud.

DEFINE_mBool(enable_stream_load_commit_txn_on_be, "false");

// The buffer size to store stream table function schema info

DEFINE_Int64(stream_tvf_buffer_size, "1048576"); // 1MB

// request cdc client timeout

DEFINE_mInt32(request_cdc_client_timeout_ms, "120000");

// OlapTableSink sender's send interval, should be less than the real response time of a tablet writer rpc.

// You may need to lower the speed when the sink receiver bes are too busy.

DEFINE_mInt32(olap_table_sink_send_interval_microseconds, "1000");

DEFINE_mDouble(olap_table_sink_send_interval_auto_partition_factor, "0.001");

// Fragment thread pool

DEFINE_Int32(fragment_mgr_async_work_pool_thread_num_min, "16");

DEFINE_Int32(fragment_mgr_async_work_pool_thread_num_max, "2048");

DEFINE_Int32(fragment_mgr_async_work_pool_queue_size, "4096");

// The read size is the size of the reads sent to os.

// There is a trade off of latency and throughout, trying to keep disks busy but

// not introduce seeks. The literature seems to agree that with 8 MB reads, random

// io and sequential io perform similarly.

DEFINE_Int32(min_buffer_size, "1024"); // 1024, The minimum read buffer size (in bytes)

// for pprof

DEFINE_String(pprof_profile_dir, "${DORIS_HOME}/log");

// for jeprofile in jemalloc

DEFINE_mString(jeprofile_dir, "${DORIS_HOME}/log");

DEFINE_mBool(enable_je_purge_dirty_pages, "true");

DEFINE_mInt32(je_dirty_decay_ms, "5000");

// to forward compatibility, will be removed later

DEFINE_mBool(enable_token_check, "true");

// to open/close system metrics

DEFINE_Bool(enable_system_metrics, "true");

// Number of cores Doris will used, this will effect only when it's greater than 0.

// Otherwise, Doris will use all cores returned from "/proc/cpuinfo".

DEFINE_Int32(num_cores, "0");

// When BE start, If there is a broken disk, BE process will exit by default.

// Otherwise, we will ignore the broken disk,

DEFINE_Bool(ignore_broken_disk, "false");

DEFINE_Bool(ignore_file_cache_dir_upgrade_failure, "false");

// Sleep time in milliseconds between memory maintenance iterations

DEFINE_mInt32(memory_maintenance_sleep_time_ms, "50");

// Memory gc are expensive, wait a while to avoid too frequent.

DEFINE_mInt32(memory_gc_sleep_time_ms, "500");

// max write buffer size before flush, default 200MB

DEFINE_mInt64(write_buffer_size, "209715200");

DEFINE_mBool(enable_adaptive_write_buffer_size, "true");

// max buffer size used in memtable for the aggregated table, default 400MB

DEFINE_mInt64(write_buffer_size_for_agg, "104857600");

DEFINE_mInt64(min_write_buffer_size_for_partial_update, "1048576");

// max parallel flush task per memtable writer

DEFINE_mInt32(memtable_flush_running_count_limit, "2");

// maximum sleep time to wait for memory when writing or flushing memtable.

DEFINE_mInt32(memtable_wait_for_memory_sleep_time_s, "300");

DEFINE_Int32(load_process_max_memory_limit_percent, "50"); // 50%

// If the memory consumption of load jobs exceed load_process_max_memory_limit,

// all load jobs will hang there to wait for memtable flush. We should have a

// soft limit which can trigger the memtable flush for the load channel who

// consumes lagest memory size before we reach the hard limit. The soft limit

// might avoid all load jobs hang at the same time.

DEFINE_Int32(load_process_soft_mem_limit_percent, "80");

// If load memory consumption is within load_process_safe_mem_permit_percent,

// memtable memory limiter will do nothing.

DEFINE_Int32(load_process_safe_mem_permit_percent, "5");

// result buffer cancelled time (unit: second)

DEFINE_mInt32(result_buffer_cancelled_interval_time, "300");

// arrow flight result sink buffer rows size, default 4096 * 8

DEFINE_mInt32(arrow_flight_result_sink_buffer_size_rows, "32768");

// The timeout for ADBC Client to wait for data using arrow flight reader.

// If the query is very complex and no result is generated after this time, consider increasing this timeout.

DEFINE_mInt32(arrow_flight_reader_brpc_controller_timeout_ms, "300000");

// the increased frequency of priority for remaining tasks in BlockingPriorityQueue

DEFINE_mInt32(priority_queue_remaining_tasks_increased_frequency, "512");

// sync tablet_meta when modifying meta

DEFINE_mBool(sync_tablet_meta, "false");

// sync when closing a file writer

DEFINE_mBool(sync_file_on_close, "true");

// default thrift rpc timeout ms

DEFINE_mInt32(thrift_rpc_timeout_ms, "60000");

// txn commit rpc timeout

DEFINE_mInt32(txn_commit_rpc_timeout_ms, "180000");

// If set to true, metric calculator will run

DEFINE_Bool(enable_metric_calculator, "true");

// max consumer num in one data consumer group, for routine load

DEFINE_mInt32(max_consumer_num_per_group, "3");

// the max size of thread pool for routine load task.

// this should be larger than FE config 'max_routine_load_task_num_per_be' (default 5)

DEFINE_Int32(max_routine_load_thread_pool_size, "1024");

// the timeout of condition variable wait in blocking_get and blocking_put

DEFINE_mInt32(blocking_queue_cv_wait_timeout_ms, "1000");

// max external scan cache batch count, means cache max_memory_cache_batch_count * batch_size row

// default is 20, batch_size's default value is 1024 means 20 * 1024 rows will be cached

DEFINE_mInt32(max_memory_sink_batch_count, "20");

// This configuration is used for the context gc thread schedule period

// note: unit is minute, default is 5min

DEFINE_mInt32(scan_context_gc_interval_min, "5");

// es scroll keep-alive

DEFINE_String(es_scroll_keepalive, "5m");

// HTTP connection timeout for es

DEFINE_mInt32(es_http_timeout_ms, "5000");

// the max client cache number per each host

// There are variety of client cache in BE, but currently we use the

// same cache size configuration.

// TODO(cmy): use different config to set different client cache if necessary.

DEFINE_Int32(max_client_cache_size_per_host, "10");

DEFINE_Int32(max_master_fe_client_cache_size, "10");

// Dir to save files downloaded by SmallFileMgr

DEFINE_String(small_file_dir, "${DORIS_HOME}/lib/small_file/");

// path gc

DEFINE_Bool(path_gc_check, "true");

DEFINE_mInt32(path_gc_check_interval_second, "86400");

DEFINE_mInt32(path_gc_check_step, "1000");

DEFINE_mInt32(path_gc_check_step_interval_ms, "10");

// The following 2 configs limit the max usage of disk capacity of a data dir.

// If both of these 2 threshold reached, no more data can be writen into that data dir.

// The percent of max used capacity of a data dir

DEFINE_mInt32(storage_flood_stage_usage_percent, "90"); // 90%

// The min bytes that should be left of a data dir

DEFINE_mInt64(storage_flood_stage_left_capacity_bytes, "1073741824"); // 1GB

// number of thread for flushing memtable per store

DEFINE_mInt32(flush_thread_num_per_store, "6");

// number of thread for flushing memtable per store, for high priority load task

DEFINE_mInt32(high_priority_flush_thread_num_per_store, "6");

// number of threads = min(flush_thread_num_per_store * num_store,

//                         max_flush_thread_num_per_cpu * num_cpu)

DEFINE_mInt32(max_flush_thread_num_per_cpu, "4");

// minimum flush threads per cpu when adaptive flush is enabled (default 0.5)

DEFINE_mDouble(min_flush_thread_num_per_cpu, "0.5");

// Whether to enable adaptive flush thread adjustment

DEFINE_mBool(enable_adaptive_flush_threads, "true");

// Whether to block writes when one table has too many pending flush memtables on this BE.

DEFINE_mBool(enable_table_memtable_flush_backpressure, "true");

// Max pending flush memtables for one table on this BE before blocking new writes.

DEFINE_mInt32(table_memtable_flush_pending_count_limit, "10");

// config for tablet meta checkpoint

DEFINE_mInt32(tablet_meta_checkpoint_min_new_rowsets_num, "10");

DEFINE_mInt32(tablet_meta_checkpoint_min_interval_secs, "600");

DEFINE_Int32(generate_tablet_meta_checkpoint_tasks_interval_secs, "600");

// config for default rowset type

// Valid configs: ALPHA, BETA

DEFINE_String(default_rowset_type, "BETA");

// Maximum size of a single message body in all protocols

DEFINE_Int64(brpc_max_body_size, "3147483648");

DEFINE_Int64(brpc_socket_max_unwritten_bytes, "-1");

DEFINE_mBool(brpc_usercode_in_pthread, "false");

// TODO(zxy): expect to be true in v1.3

// Whether to embed the ProtoBuf Request serialized string together with Tuple/Block data into

// Controller Attachment and send it through http brpc when the length of the Tuple/Block data

// is greater than 1.8G. This is to avoid the error of Request length overflow (2G).

DEFINE_mBool(transfer_large_data_by_brpc, "true");

// max number of txns for every txn_partition_map in txn manager

// this is a self protection to avoid too many txns saving in manager

DEFINE_mInt64(max_runnings_transactions_per_txn_map, "2000");

// tablet_map_lock shard size, the value is 2^n, n=0,1,2,3,4

// this is a an enhancement for better performance to manage tablet

DEFINE_Int32(tablet_map_shard_size, "256");

// txn_map_lock shard size, the value is 2^n, n=0,1,2,3,4

// this is a an enhancement for better performance to manage txn

DEFINE_Int32(txn_map_shard_size, "1024");

// txn_lock shard size, the value is 2^n, n=0,1,2,3,4

// this is a an enhancement for better performance to commit and publish txn

DEFINE_Int32(txn_shard_size, "1024");

// Whether to continue to start be when load tablet from header failed.

DEFINE_Bool(ignore_load_tablet_failure, "true");

// Whether to continue to start be when load tablet from header failed.

DEFINE_mBool(ignore_rowset_stale_unconsistent_delete, "false");

// Set max cache's size of query results, the unit is M byte

DEFINE_Int32(query_cache_max_size_mb, "256");

// Cache memory is pruned when reach query_cache_max_size_mb + query_cache_elasticity_size_mb

DEFINE_Int32(query_cache_elasticity_size_mb, "128");

// Maximum number of cache partitions corresponding to a SQL

DEFINE_Int32(query_cache_max_partition_count, "1024");

// Maximum number of version of a tablet. If the version num of a tablet exceed limit,

// the load process will reject new incoming load job of this tablet.

// This is to avoid too many version num.

DEFINE_mInt32(max_tablet_version_num, "2000");

DEFINE_mInt32(time_series_max_tablet_version_num, "20000");

// the max sleep time when meeting high pressure load task

DEFINE_mInt64(max_load_back_pressure_version_wait_time_ms, "3000");

// the threshold of rowset number gap that triggers back pressure

DEFINE_mInt64(load_back_pressure_version_threshold, "80"); // 80%

// Frontend mainly use two thrift sever type: THREAD_POOL, THREADED_SELECTOR. if fe use THREADED_SELECTOR model for thrift server,

// the thrift_server_type_of_fe should be set THREADED_SELECTOR to make be thrift client to fe constructed with TFramedTransport

DEFINE_String(thrift_server_type_of_fe, "THREAD_POOL");

// disable zone map index when page row is too few

DEFINE_mInt32(zone_map_row_num_threshold, "20");

// Maximum number of IN values checked exactly against a zone map. For larger sets, only the

// IN-set min/max range is checked.

DEFINE_mInt32(in_zonemap_point_check_threshold, "8192");

// aws sdk log level

//    Off = 0,

//    Fatal = 1,

//    Error = 2,

//    Warn = 3,

//    Info = 4,

//    Debug = 5,

//    Trace = 6

DEFINE_Int32(aws_log_level, "2");

DEFINE_Validator(aws_log_level,

                 [](const int config) -> bool { return config >= 0 && config <= 6; });

// azure sdk log level

//    Verbose = 1,

//    Informational = 2,

//    Warning = 3,

//    Error = 4

DEFINE_Int32(azure_log_level, "4");

DEFINE_Validator(azure_log_level,

                 [](const int config) -> bool { return config >= 1 && config <= 4; });

// the buffer size when read data from remote storage like s3

DEFINE_mInt32(remote_storage_read_buffer_mb, "16");

// The minimum length when TCMalloc Hook consumes/releases MemTracker, consume size

// smaller than this value will continue to accumulate. specified as number of bytes.

// Decreasing this value will increase the frequency of consume/release.

// Increasing this value will cause MemTracker statistics to be inaccurate.

DEFINE_mInt32(mem_tracker_consume_min_size_bytes, "1048576");

// The version information of the tablet will be stored in the memory

// in an adjacency graph data structure.

// And as the new version is written and the old version is deleted,

// the data structure will begin to have empty vertex with no edge associations(orphan vertex).

// This config is used to control that when the proportion of orphan vertex is greater than the threshold,

// the adjacency graph will be rebuilt to ensure that the data structure will not expand indefinitely.

// This config usually only needs to be modified during testing.

// In most cases, it does not need to be modified.

DEFINE_mDouble(tablet_version_graph_orphan_vertex_ratio, "0.1");

// share delta writers when memtable_on_sink_node = true

DEFINE_Bool(share_delta_writers, "true");

// timeout for open load stream rpc in ms

DEFINE_Int64(open_load_stream_timeout_ms, "60000"); // 60s

// enable write background when using brpc stream

DEFINE_mBool(enable_brpc_stream_write_background, "true");

// brpc streaming max_buf_size in bytes

DEFINE_Int64(load_stream_max_buf_size, "20971520"); // 20MB

// brpc streaming messages_in_batch

DEFINE_Int32(load_stream_messages_in_batch, "128");

// brpc streaming StreamWait seconds on EAGAIN

DEFINE_Int32(load_stream_eagain_wait_seconds, "600");

// max tasks per flush token in load stream

DEFINE_Int32(load_stream_flush_token_max_tasks, "15");

// max wait flush token time in load stream

DEFINE_Int32(load_stream_max_wait_flush_token_time_ms, "600000");

// number of send batch thread pool size

DEFINE_Int32(send_batch_thread_pool_thread_num, "64");

// number of send batch thread pool queue size

DEFINE_Int32(send_batch_thread_pool_queue_size, "102400");

// Limit the number of segment of a newly created rowset.

// The newly created rowset may to be compacted after loading,

// so if there are too many segment in a rowset, the compaction process

// will run out of memory.

// When doing compaction, each segment may take at least 1MB buffer.

DEFINE_mInt32(max_segment_num_per_rowset, "1000");

DEFINE_mInt32(segment_compression_threshold_kb, "256");

// Global bitmap cache capacity for aggregation cache, size in bytes

DEFINE_Int64(delete_bitmap_agg_cache_capacity, "104857600");

// The default delete bitmap cache is set to 100MB,

// which can be insufficient and cause performance issues when the amount of user data is large.

// To mitigate the problem of an inadequate cache,

// we will take the larger of 1.0% of the total memory and 100MB as the delete bitmap cache size.

DEFINE_String(delete_bitmap_dynamic_agg_cache_limit, "1.0%");

DEFINE_mInt32(delete_bitmap_agg_cache_stale_sweep_time_sec, "1800");

// reference https://github.com/edenhill/librdkafka/blob/master/INTRODUCTION.md#broker-version-compatibility

// If the dependent kafka broker version older than 0.10.0.0,

// the value of kafka_api_version_request should be false, and the

// value set by the fallback version kafka_broker_version_fallback will be used,

// and the valid values are: 0.9.0.x, 0.8.x.y.

DEFINE_String(kafka_api_version_request, "true");

DEFINE_String(kafka_broker_version_fallback, "0.10.0");

DEFINE_String(kafka_debug, "disable");

// The number of pool siz of routine load consumer.

// If you meet the error describe in https://github.com/edenhill/librdkafka/issues/3608

// Change this size to 0 to fix it temporarily.

DEFINE_mInt32(routine_load_consumer_pool_size, "1024");

// Used in single-stream-multi-table load. When receive a batch of messages from kafka,

// if the size of batch is more than this threshold, we will request plans for all related tables.

DEFINE_Int32(multi_table_batch_plan_threshold, "200");

// Used in single-stream-multi-table load. When receiving a batch of messages from Kafka,

// if the size of the table wait for plan is more than this threshold, we will request plans for all related tables.

// The param is aimed to avoid requesting and executing too many plans at once.

// Performing small batch processing on multiple tables during the loaded process can reduce the pressure of a single RPC

// and improve the real-time processing of data.

DEFINE_Int32(multi_table_max_wait_tables, "5");

// When the timeout of a load task is less than this threshold,

// Doris treats it as a high priority task.

// high priority tasks use a separate thread pool for flush and do not block rpc by memory cleanup logic.

// this threshold is mainly used to identify routine load tasks and should not be modified if not necessary.

DEFINE_mInt32(load_task_high_priority_threshold_second, "600");

// The min timeout of load rpc (add batch, close, etc.)

// Because a load rpc may be blocked for a while.

// Increase this config may avoid rpc timeout.

DEFINE_mInt32(min_load_rpc_timeout_ms, "20000");

// use which protocol to access function service, candicate is baidu_std/h2:grpc

DEFINE_String(function_service_protocol, "h2:grpc");

// use which load balancer to select server to connect

DEFINE_String(rpc_load_balancer, "rr");

// a soft limit of string type length, the hard limit is 2GB - 4, but if too long will cause very low performance,

// so we set a soft limit, default is 10MB

DEFINE_Int32(string_type_length_soft_limit_bytes, "10485760");

DEFINE_Validator(string_type_length_soft_limit_bytes,

                 [](const int config) -> bool { return config > 0 && config <= 2147483643; });

// Threshold of reading a small file into memory

DEFINE_mInt32(in_memory_file_size, "1048576"); // 1MB

// Max size of parquet page header in bytes

DEFINE_mInt32(parquet_header_max_size_mb, "1");

// Max size of parquet file metadata in bytes

DEFINE_mInt64(parquet_metadata_size_limit, "268435456");

DEFINE_Validator(parquet_metadata_size_limit,

                 [](const int64_t config) -> bool { return config > 0; });

// Max buffer size for parquet row group

DEFINE_mInt32(parquet_rowgroup_max_buffer_mb, "128");

// Max buffer size for parquet chunk column

DEFINE_mInt32(parquet_column_max_buffer_mb, "8");

DEFINE_mDouble(max_amplified_read_ratio, "0.8");

DEFINE_mInt32(merged_oss_min_io_size, "1048576");

DEFINE_mInt32(merged_hdfs_min_io_size, "8192");

// OrcReader

DEFINE_mInt32(orc_natural_read_size_mb, "8");

DEFINE_Validator(orc_natural_read_size_mb,

                 [](const int config) -> bool { return config > 0 && config <= 1024; });

// Perform the always_true check at intervals determined by runtime_filter_sampling_frequency

DEFINE_mInt32(runtime_filter_sampling_frequency, "32");

DEFINE_mInt32(execution_max_rpc_timeout_sec, "3600");

DEFINE_mBool(execution_ignore_eovercrowded, "true");

// cooldown task configs

DEFINE_Int32(cooldown_thread_num, "5");

DEFINE_mInt64(generate_cooldown_task_interval_sec, "20");

DEFINE_mInt32(remove_unused_remote_files_interval_sec, "21600"); // 6h

DEFINE_mInt32(confirm_unused_remote_files_interval_sec, "60");

DEFINE_Int32(cold_data_compaction_thread_num, "2");

DEFINE_mInt32(cold_data_compaction_interval_sec, "1800");

DEFINE_mInt32(cold_data_compaction_score_threshold, "100");

DEFINE_String(tmp_file_dir, "tmp");

DEFINE_Int32(min_s3_file_system_thread_num, "16");

DEFINE_Int32(max_s3_file_system_thread_num, "64");