Skip to content

Commit

Permalink
[fix](scanner) Fix incorrect _max_thread_num in scanner context (apac…
Browse files Browse the repository at this point in the history 
…he#40569)

1. Minor refactor for scanner constructor, calculation of
_max_thread_num is moved to init method
2. The expected value of _max_thread_num is changed. There is no need to
submit too many scan task to scan scheduler, since thread num is
limited.
3. Calculation of _max_bytes_in_queue is changed. _max_bytes_in_queue
for each scan instance is limited to 100MB by default.
  • Loading branch information
@zhiqiang-hhhh @dataroaring
zhiqiang-hhhh authored and dataroaring committed Oct 5, 2024
1 parent 7140f9a commit 0be00cc
Show file tree
Hide file tree
Showing 4 changed files with 79 additions and 66 deletions.
11 changes: 1 addition & 10 deletions be/src/pipeline/exec/scan_operator.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -994,16 +994,7 @@ Status ScanLocalState<Derived>::_start_scanners(
auto& p = _parent->cast<typename Derived::Parent>();
_scanner_ctx = vectorized::ScannerContext::create_shared(
state(), this, p._output_tuple_desc, p.output_row_descriptor(), scanners, p.limit(),
state()->scan_queue_mem_limit(), _scan_dependency,
// NOTE: This will logic makes _max_thread_num of ScannerContext to be C(num of cores) * 2
// For a query with C/2 instance and M scan node, scan task of this query will be C/2 * M * C*2
// and will be C*C*N at most.
// 1. If data distribution is ignored , we use 1 instance to scan.
// 2. Else if this operator is not file scan operator, we use config::doris_scanner_thread_pool_thread_num scanners to scan.
// 3. Else, file scanner will consume much memory so we use config::doris_scanner_thread_pool_thread_num / query_parallel_instance_num scanners to scan.
p.ignore_data_distribution() || !p.is_file_scan_operator()
? 1
: state()->query_parallel_instance_num());
_scan_dependency, p.ignore_data_distribution());
return Status::OK();
}

Expand Down
120 changes: 70 additions & 50 deletions be/src/vec/exec/scan/scanner_context.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -42,8 +42,7 @@ ScannerContext::ScannerContext(
RuntimeState* state, pipeline::ScanLocalStateBase* local_state,
const TupleDescriptor* output_tuple_desc, const RowDescriptor* output_row_descriptor,
const std::list<std::shared_ptr<vectorized::ScannerDelegate>>& scanners, int64_t limit_,
int64_t max_bytes_in_blocks_queue, std::shared_ptr<pipeline::Dependency> dependency,
const int num_parallel_instances)
std::shared_ptr<pipeline::Dependency> dependency, bool ignore_data_distribution)
: HasTaskExecutionCtx(state),
_state(state),
_local_state(local_state),
Expand All @@ -53,53 +52,102 @@ ScannerContext::ScannerContext(
_output_row_descriptor(output_row_descriptor),
_batch_size(state->batch_size()),
limit(limit_),
_max_bytes_in_queue(std::max(max_bytes_in_blocks_queue, (int64_t)1024) *
num_parallel_instances),
_scanner_scheduler(state->exec_env()->scanner_scheduler()),
_all_scanners(scanners.begin(), scanners.end()),
_num_parallel_instances(num_parallel_instances) {
_ignore_data_distribution(ignore_data_distribution) {
DCHECK(_output_row_descriptor == nullptr ||
_output_row_descriptor->tuple_descriptors().size() == 1);
_query_id = _state->get_query_ctx()->query_id();
ctx_id = UniqueId::gen_uid().to_string();
// Provide more memory for wide tables, increase proportionally by multiples of 300
_max_bytes_in_queue *= _output_tuple_desc->slots().size() / 300 + 1;
if (scanners.empty()) {
_is_finished = true;
_set_scanner_done();
}
_scanners.enqueue_bulk(scanners.begin(), scanners.size());
if (limit < 0) {
limit = -1;
}
MAX_SCALE_UP_RATIO = _state->scanner_scale_up_ratio();
_query_thread_context = {_query_id, _state->query_mem_tracker(),
_state->get_query_ctx()->workload_group()};
_dependency = dependency;
}

// After init function call, should not access _parent
Status ScannerContext::init() {
_scanner_profile = _local_state->_scanner_profile;
_scanner_sched_counter = _local_state->_scanner_sched_counter;
_newly_create_free_blocks_num = _local_state->_newly_create_free_blocks_num;
_scanner_wait_batch_timer = _local_state->_scanner_wait_batch_timer;
_scanner_ctx_sched_time = _local_state->_scanner_ctx_sched_time;
_scale_up_scanners_counter = _local_state->_scale_up_scanners_counter;

#ifndef BE_TEST
// 3. get thread token
if (_state->get_query_ctx()) {
thread_token = _state->get_query_ctx()->get_token();
_simple_scan_scheduler = _state->get_query_ctx()->get_scan_scheduler();
if (_simple_scan_scheduler) {
_should_reset_thread_name = false;
}
_remote_scan_task_scheduler = _state->get_query_ctx()->get_remote_scan_scheduler();
}
#endif
_local_state->_runtime_profile->add_info_string("UseSpecificThreadToken",
thread_token == nullptr ? "False" : "True");

int num_parallel_instances = _state->query_parallel_instance_num();

// NOTE: When ignore_data_distribution is true, the parallelism
// of the scan operator is regarded as 1 (actually maybe not).
// That will make the number of scan task can be submitted to the scheduler
// in a vary large value. This logicl is kept from the older implementation.
// https://github.com/apache/doris/pull/28266
if (_ignore_data_distribution) {
num_parallel_instances = 1;
}

// _max_bytes_in_queue controls the maximum memory that can be used by a single scan instance.
// scan_queue_mem_limit on FE is 100MB by default, on backend we will make sure its actual value
// is larger than 10MB.
_max_bytes_in_queue = std::max(_state->scan_queue_mem_limit(), (int64_t)1024 * 1024 * 10);

// Provide more memory for wide tables, increase proportionally by multiples of 300
_max_bytes_in_queue *= _output_tuple_desc->slots().size() / 300 + 1;

// TODO: Where is the proper position to place this code?
if (_all_scanners.empty()) {
_is_finished = true;
_set_scanner_done();
}

// _max_thread_num controls how many scanners of this ScanOperator can be submitted to scheduler at a time.
// The overall target of our system is to make full utilization of the resources.
// At the same time, we dont want too many tasks are queued by scheduler, that makes the query
// waiting too long, and existing task can not be scheduled in time.
// First of all, we try to make sure _max_thread_num of a ScanNode of a query on a single backend is less than
// config::doris_scanner_thread_pool_thread_num.
// At the same time, we dont want too many tasks are queued by scheduler, that is not necessary.
// So, first of all, we try to make sure _max_thread_num of a ScanNode of a query on a single backend is less than
// 2 * config::doris_scanner_thread_pool_thread_num, so that we can make all io threads busy.
// For example, on a 64-core machine, the default value of config::doris_scanner_thread_pool_thread_num will be 64*2 =128.
// and the num_parallel_instances of this scan operator will be 64/2=32.
// For a query who has two scan nodes, the _max_thread_num of each scan node instance will be 128 / 32 = 4.
// We have 32 instances of this scan operator, so for the ScanNode, we have 4 * 32 = 128 scanner tasks can be submitted at a time.
// Remember that we have to ScanNode in this query, so the total number of scanner tasks can be submitted at a time is 128 * 2 = 256.
// For a query who has one scan nodes, the _max_thread_num of each scan node instance will be 2 * 128 / 32 = 8.
// We have 32 instances of this scan operator, so for the ScanNode, we have 8 * 32 = 256 scanner tasks can be submitted at a time.
// The thread pool of scanner is 128, that means we will have 128 tasks running in parallel and another 128 tasks are waiting in the queue.
// When first 128 tasks are finished, the next 128 tasks will be extricated from the queue and be executed,
// and another 128 tasks will be submitted to the queue if there are remaining.
_max_thread_num =
_state->num_scanner_threads() > 0
? _state->num_scanner_threads()
: config::doris_scanner_thread_pool_thread_num / num_parallel_instances;
: 2 * (config::doris_scanner_thread_pool_thread_num / num_parallel_instances);
_max_thread_num = _max_thread_num == 0 ? 1 : _max_thread_num;
// In some situation, there are not too many big tablets involed, so we can reduce the thread number.
_max_thread_num = std::min(_max_thread_num, (int32_t)scanners.size());
// NOTE: when _all_scanners.size is zero, the _max_thread_num will be 0.
_max_thread_num = std::min(_max_thread_num, (int32_t)_all_scanners.size());

// 1. Calculate max concurrency
// For select * from table limit 10; should just use one thread.
if (_local_state->should_run_serial()) {
_max_thread_num = 1;
}

// when user not specify scan_thread_num, so we can try downgrade _max_thread_num.
// becaue we found in a table with 5k columns, column reader may ocuppy too much memory.
// you can refer https://github.com/apache/doris/issues/35340 for details.
int32_t max_column_reader_num = state->query_options().max_column_reader_num;
int32_t max_column_reader_num = _state->query_options().max_column_reader_num;
if (_max_thread_num != 1 && max_column_reader_num > 0) {
int32_t scan_column_num = _output_tuple_desc->slots().size();
int32_t current_column_num = scan_column_num * _max_thread_num;
Expand All @@ -109,43 +157,15 @@ ScannerContext::ScannerContext(
if (new_max_thread_num < _max_thread_num) {
int32_t origin_max_thread_num = _max_thread_num;
_max_thread_num = new_max_thread_num;
LOG(INFO) << "downgrade query:" << print_id(state->query_id())
LOG(INFO) << "downgrade query:" << print_id(_state->query_id())
<< " scan's max_thread_num from " << origin_max_thread_num << " to "
<< _max_thread_num << ",column num: " << scan_column_num
<< ", max_column_reader_num: " << max_column_reader_num;
}
}
}

_query_thread_context = {_query_id, _state->query_mem_tracker(),
_state->get_query_ctx()->workload_group()};
_dependency = dependency;
}

// After init function call, should not access _parent
Status ScannerContext::init() {
_scanner_profile = _local_state->_scanner_profile;
_scanner_sched_counter = _local_state->_scanner_sched_counter;
_newly_create_free_blocks_num = _local_state->_newly_create_free_blocks_num;
_scanner_wait_batch_timer = _local_state->_scanner_wait_batch_timer;
_scanner_ctx_sched_time = _local_state->_scanner_ctx_sched_time;
_scale_up_scanners_counter = _local_state->_scale_up_scanners_counter;

#ifndef BE_TEST
// 3. get thread token
if (_state->get_query_ctx()) {
thread_token = _state->get_query_ctx()->get_token();
_simple_scan_scheduler = _state->get_query_ctx()->get_scan_scheduler();
if (_simple_scan_scheduler) {
_should_reset_thread_name = false;
}
_remote_scan_task_scheduler = _state->get_query_ctx()->get_remote_scan_scheduler();
}
#endif

COUNTER_SET(_local_state->_max_scanner_thread_num, (int64_t)_max_thread_num);
_local_state->_runtime_profile->add_info_string("UseSpecificThreadToken",
thread_token == nullptr ? "False" : "True");

// submit `_max_thread_num` running scanners to `ScannerScheduler`
// When a running scanners is finished, it will submit one of the remaining scanners.
Expand Down
9 changes: 4 additions & 5 deletions be/src/vec/exec/scan/scanner_context.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -105,9 +105,8 @@ class ScannerContext : public std::enable_shared_from_this<ScannerContext>,
const TupleDescriptor* output_tuple_desc,
const RowDescriptor* output_row_descriptor,
const std::list<std::shared_ptr<vectorized::ScannerDelegate>>& scanners,
int64_t limit_, int64_t max_bytes_in_blocks_queue,
std::shared_ptr<pipeline::Dependency> dependency,
const int num_parallel_instances);
int64_t limit_, std::shared_ptr<pipeline::Dependency> dependency,
bool ignore_data_distribution);

~ScannerContext() override {
SCOPED_SWITCH_THREAD_MEM_TRACKER_LIMITER(_query_thread_context.query_mem_tracker);
Expand Down Expand Up @@ -210,7 +209,7 @@ class ScannerContext : public std::enable_shared_from_this<ScannerContext>,
int64_t limit;

int32_t _max_thread_num = 0;
int64_t _max_bytes_in_queue;
int64_t _max_bytes_in_queue = 0;
doris::vectorized::ScannerScheduler* _scanner_scheduler;
SimplifiedScanScheduler* _simple_scan_scheduler = nullptr;
SimplifiedScanScheduler* _remote_scan_task_scheduler = nullptr;
Expand All @@ -220,7 +219,6 @@ class ScannerContext : public std::enable_shared_from_this<ScannerContext>,
int32_t _num_running_scanners = 0;
// weak pointer for _scanners, used in stop function
std::vector<std::weak_ptr<ScannerDelegate>> _all_scanners;
const int _num_parallel_instances;
std::shared_ptr<RuntimeProfile> _scanner_profile;
RuntimeProfile::Counter* _scanner_sched_counter = nullptr;
RuntimeProfile::Counter* _newly_create_free_blocks_num = nullptr;
Expand All @@ -229,6 +227,7 @@ class ScannerContext : public std::enable_shared_from_this<ScannerContext>,
RuntimeProfile::Counter* _scale_up_scanners_counter = nullptr;
QueryThreadContext _query_thread_context;
std::shared_ptr<pipeline::Dependency> _dependency = nullptr;
bool _ignore_data_distribution = false;

// for scaling up the running scanners
size_t _estimated_block_size = 0;
Expand Down
5 changes: 4 additions & 1 deletion fe/fe-core/src/main/java/org/apache/doris/qe/SessionVariable.java
View file
Original file line number Diff line number Diff line change
Expand Up @@ -682,7 +682,10 @@ public class SessionVariable implements Serializable, Writable {
@VariableMgr.VarAttr(name = EXEC_MEM_LIMIT)
public long maxExecMemByte = 2147483648L;

@VariableMgr.VarAttr(name = SCAN_QUEUE_MEM_LIMIT)
@VariableMgr.VarAttr(name = SCAN_QUEUE_MEM_LIMIT,
description = {"每个 Scan Instance 的 block queue 能够保存多少字节的 block",
"How many bytes of block can be saved in the block queue of each Scan Instance"})
// 100MB
public long maxScanQueueMemByte = 2147483648L / 20;

@VariableMgr.VarAttr(name = NUM_SCANNER_THREADS, needForward = true, description = {
Expand Down

0 comments on commit 0be00cc

Please sign in to comment.