pengxiangyu commented on code in PR #10280:
URL: https://github.com/apache/doris/pull/10280#discussion_r911892938
##########
be/src/agent/agent_server.cpp:
##########
@@ -153,8 +154,8 @@ void AgentServer::submit_tasks(TAgentResult& agent_result,
HANDLE_TYPE(TTaskType::UPDATE_TABLET_META_INFO,
_update_tablet_meta_info_workers,
update_tablet_meta_info_req);
HANDLE_TYPE(TTaskType::COMPACTION,
_submit_table_compaction_workers, compaction_req);
- HANDLE_TYPE(TTaskType::STORAGE_MEDIUM_MIGRATE_V2,
_storage_medium_migrate_v2_workers,
- storage_migration_req_v2);
Review Comment:
where is _storage_refresh_policy_workers? is it not needed here?
##########
be/src/agent/task_worker_pool.cpp:
##########
@@ -1683,9 +1701,64 @@ void
TaskWorkerPool::_submit_table_compaction_worker_thread_callback() {
}
}
-void TaskWorkerPool::_storage_medium_migrate_v2_worker_thread_callback() {
+void TaskWorkerPool::_storage_refresh_storage_policy_worker_thread_callback() {
+ while (_is_work) {
+ _is_doing_work = false;
+ // wait at most report_task_interval_seconds, or being notified
+ std::unique_lock<std::mutex> worker_thread_lock(_worker_thread_lock);
+ _worker_thread_condition_variable.wait_for(
+ worker_thread_lock,
+
std::chrono::seconds(config::storage_refresh_storage_policy_task_interval_seconds));
+ if (!_is_work) {
+ break;
+ }
+
+ if (_master_info.network_address.port == 0) {
+ // port == 0 means not received heartbeat yet
+ // sleep a short time and try again
+ LOG(INFO)
+ << "waiting to receive first heartbeat from frontend
before doing task report";
+ continue;
+ }
+
+ _is_doing_work = true;
+
+ TGetStoragePolicyResult result;
+ Status status = _master_client->refresh_storage_policy(&result);
+ if (!status.ok()) {
+ LOG(WARNING) << "refresh storage policy status not ok. ";
+ } else if (result.status.status_code != TStatusCode::OK) {
+ LOG(WARNING) << "refresh storage policy result status status_code
not ok. ";
Review Comment:
unexpect space after 'status_code not ok.'
##########
be/src/olap/rowset/beta_rowset_writer.cpp:
##########
@@ -304,29 +270,28 @@ RowsetSharedPtr BetaRowsetWriter::build() {
Status BetaRowsetWriter::_create_segment_writer(
std::unique_ptr<segment_v2::SegmentWriter>* writer) {
- auto path_desc =
- BetaRowset::segment_file_path(_context.path_desc,
_context.rowset_id, _num_segment++);
- // TODO(lingbin): should use a more general way to get BlockManager object
- // and tablets with the same type should share one BlockManager object;
- fs::BlockManager* block_mgr =
fs::fs_util::block_manager(_context.path_desc);
- std::unique_ptr<fs::WritableBlock> wblock;
- fs::CreateBlockOptions opts(path_desc);
- DCHECK(block_mgr != nullptr);
- Status st = block_mgr->create_block(opts, &wblock);
+ auto path = BetaRowset::local_segment_path(_context.tablet_path,
_context.rowset_id,
+ _num_segment++);
+ auto fs = _rowset_meta->fs();
+ if (!fs) {
Review Comment:
ERROR LOG needed.
##########
be/src/olap/base_compaction.cpp:
##########
@@ -106,14 +106,11 @@ Status BaseCompaction::pick_rowsets_to_compact() {
}
// 2. the ratio between base rowset and all input cumulative rowsets
reaches the threshold
- int64_t base_size = 0;
+ // `_input_rowsets` has been sorted by end version, so we consider
`_input_rowsets[0]` is the base rowset.
+ int64_t base_size = _input_rowsets.front()->data_disk_size();
int64_t cumulative_total_size = 0;
- for (auto& rowset : _input_rowsets) {
- if (rowset->start_version() != 0) {
- cumulative_total_size += rowset->data_disk_size();
- } else {
- base_size = rowset->data_disk_size();
- }
+ for (auto it = _input_rowsets.begin() + 1; it != _input_rowsets.end();
++it) {
Review Comment:
_input_rowsets may be not in order, check it is better.
##########
be/src/olap/tablet.cpp:
##########
@@ -1636,4 +1652,160 @@ std::shared_ptr<MemTracker>&
Tablet::get_compaction_mem_tracker(CompactionType c
}
}
+Status Tablet::cooldown() {
+ std::unique_lock schema_change_lock(_schema_change_lock, std::try_to_lock);
+ if (!schema_change_lock.owns_lock()) {
+ LOG(WARNING) << "schema change is running. tablet=" << tablet_id();
+ return Status::OLAPInternalError(OLAP_ERR_BE_TRY_BE_LOCK_ERROR);
+ }
+ // Check executing serially with compaction task.
+ std::unique_lock base_compaction_lock(_base_compaction_lock,
std::try_to_lock);
+ if (!base_compaction_lock.owns_lock()) {
+ LOG(WARNING) << "base compaction is running. tablet=" << tablet_id();
+ return Status::OLAPInternalError(OLAP_ERR_BE_TRY_BE_LOCK_ERROR);
+ }
+ std::unique_lock cumu_compaction_lock(_cumulative_compaction_lock,
std::try_to_lock);
+ if (!cumu_compaction_lock.owns_lock()) {
+ LOG(WARNING) << "cumulative compaction is running. tablet=" <<
tablet_id();
+ return Status::OLAPInternalError(OLAP_ERR_BE_TRY_BE_LOCK_ERROR);
+ }
+ auto dest_fs = io::FileSystemMap::instance()->get(cooldown_resource());
+ if (!dest_fs) {
+ return Status::OLAPInternalError(OLAP_ERR_NOT_INITED);
+ }
+ DCHECK(dest_fs->type() == io::FileSystemType::S3);
+ auto old_rowset = pick_cooldown_rowset();
+ if (!old_rowset) {
+ LOG(WARNING) << "Cannot pick cooldown rowset in tablet " <<
tablet_id();
+ return Status::OK();
+ }
+ RowsetId new_rowset_id = StorageEngine::instance()->next_rowset_id();
+
+ auto start = std::chrono::steady_clock::now();
+
+
RETURN_IF_ERROR(old_rowset->upload_to(reinterpret_cast<io::RemoteFileSystem*>(dest_fs.get()),
+ new_rowset_id));
+
+ auto duration =
std::chrono::duration<float>(std::chrono::steady_clock::now() - start);
+ LOG(INFO) << "Upload rowset " << old_rowset->version() << " " <<
new_rowset_id.to_string()
+ << " to " << dest_fs->root_path().native() << ", tablet_id=" <<
tablet_id()
+ << ", duration=" << duration.count() << ", capacity=" <<
old_rowset->data_disk_size()
+ << ", tp=" << old_rowset->data_disk_size() / duration.count();
+
+ // gen a new rowset
+ auto new_rowset_meta =
std::make_shared<RowsetMeta>(*old_rowset->rowset_meta());
+ new_rowset_meta->set_rowset_id(new_rowset_id);
+ new_rowset_meta->set_resource_id(dest_fs->resource_id());
+ new_rowset_meta->set_fs(dest_fs);
+ new_rowset_meta->set_creation_time(time(nullptr));
+ RowsetSharedPtr new_rowset;
+ RowsetFactory::create_rowset(&_schema, _tablet_path,
std::move(new_rowset_meta), &new_rowset);
+
+ std::vector to_add {std::move(new_rowset)};
+ std::vector to_delete {std::move(old_rowset)};
+
+ std::unique_lock meta_wlock(_meta_lock);
+ modify_rowsets(to_add, to_delete);
+ save_meta();
+ return Status::OK();
+}
+
+RowsetSharedPtr Tablet::pick_cooldown_rowset() {
+ RowsetSharedPtr rowset;
+ {
+ std::shared_lock meta_rlock(_meta_lock);
+
+ // We pick the rowset with smallest start version in local.
+ int64_t smallest_version = std::numeric_limits<int64_t>::max();
+ for (const auto& it : _rs_version_map) {
+ auto& rs = it.second;
+ if (rs->is_local() && rs->start_version() < smallest_version) {
+ smallest_version = rs->start_version();
+ rowset = rs;
+ }
+ }
+ }
+ return rowset;
+}
+
+bool Tablet::need_cooldown(int64_t* cooldown_timestamp, size_t* file_size) {
+ // std::shared_lock meta_rlock(_meta_lock);
+ if (cooldown_resource().empty()) {
Review Comment:
cooldown_resource()? isn't it policy_name?
##########
be/src/olap/tablet.cpp:
##########
@@ -1636,4 +1652,160 @@ std::shared_ptr<MemTracker>&
Tablet::get_compaction_mem_tracker(CompactionType c
}
}
+Status Tablet::cooldown() {
+ std::unique_lock schema_change_lock(_schema_change_lock, std::try_to_lock);
+ if (!schema_change_lock.owns_lock()) {
+ LOG(WARNING) << "schema change is running. tablet=" << tablet_id();
+ return Status::OLAPInternalError(OLAP_ERR_BE_TRY_BE_LOCK_ERROR);
+ }
+ // Check executing serially with compaction task.
+ std::unique_lock base_compaction_lock(_base_compaction_lock,
std::try_to_lock);
+ if (!base_compaction_lock.owns_lock()) {
+ LOG(WARNING) << "base compaction is running. tablet=" << tablet_id();
+ return Status::OLAPInternalError(OLAP_ERR_BE_TRY_BE_LOCK_ERROR);
+ }
+ std::unique_lock cumu_compaction_lock(_cumulative_compaction_lock,
std::try_to_lock);
+ if (!cumu_compaction_lock.owns_lock()) {
+ LOG(WARNING) << "cumulative compaction is running. tablet=" <<
tablet_id();
+ return Status::OLAPInternalError(OLAP_ERR_BE_TRY_BE_LOCK_ERROR);
+ }
+ auto dest_fs = io::FileSystemMap::instance()->get(cooldown_resource());
+ if (!dest_fs) {
+ return Status::OLAPInternalError(OLAP_ERR_NOT_INITED);
+ }
+ DCHECK(dest_fs->type() == io::FileSystemType::S3);
+ auto old_rowset = pick_cooldown_rowset();
Review Comment:
why only one rowset? how abort the others?
##########
be/src/olap/tablet_manager.cpp:
##########
@@ -1345,18 +1271,55 @@ void
TabletManager::get_tablets_distribution_on_different_disks(
Status TabletManager::_get_storage_param(DataDir* data_dir, const std::string&
storage_name,
StorageParamPB* storage_param) {
- if (data_dir->is_remote()) {
- RETURN_WITH_WARN_IF_ERROR(
- StorageBackendMgr::instance()->get_storage_param(storage_name,
storage_param),
- Status::OLAPInternalError(OLAP_ERR_OTHER_ERROR),
- "get_storage_param failed for storage_name: " + storage_name);
- } else {
- storage_param->set_storage_medium(
-
fs::fs_util::get_storage_medium_pb(data_dir->storage_medium()));
- }
+ storage_param->set_storage_medium(
+ fs::fs_util::get_storage_medium_pb(data_dir->storage_medium()));
return Status::OK();
}
+struct SortCtx {
+ SortCtx(TabletSharedPtr tablet, int64_t cooldown_timestamp, int64_t
file_size)
+ : tablet(tablet), cooldown_timestamp(cooldown_timestamp),
file_size(file_size) {}
+ TabletSharedPtr tablet;
+ int64_t cooldown_timestamp;
+ int64_t file_size;
+};
+
+void TabletManager::get_cooldwon_tablets(std::vector<TabletSharedPtr>*
tablets) {
Review Comment:
cooldown, misspelling
##########
be/src/olap/storage_policy_mgr.cpp:
##########
@@ -0,0 +1,91 @@
+// 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 "olap/storage_policy_mgr.h"
+
+#include "io/fs/file_system.h"
+#include "io/fs/file_system_map.h"
+#include "io/fs/s3_file_system.h"
+#include "util/s3_util.h"
+
+namespace doris {
+
+void StoragePolicyMgr::update(const std::string& name, StoragePolicyPtr
policy) {
+ std::lock_guard<std::mutex> l(_mutex);
+ auto it = _policy_map.find(name);
+ if (it != _policy_map.end()) {
+ // just support change ak, sk, cooldown_ttl, cooldown_datetime
+ LOG(INFO) << "update storage policy name: " << name;
+ auto s3_fs = std::dynamic_pointer_cast<io::S3FileSystem>(
+ io::FileSystemMap::instance()->get(name));
+ DCHECK(s3_fs);
+ s3_fs->set_ak(policy->s3_ak);
+ s3_fs->set_sk(policy->s3_sk);
+ s3_fs->connect();
+ it->second = std::move(policy);
+ } else {
+ // can't find name's policy, so do nothing.
+ }
+}
+
+void StoragePolicyMgr::periodic_put(const std::string& name, StoragePolicyPtr
policy) {
+ std::lock_guard<std::mutex> l(_mutex);
+ auto it = _policy_map.find(name);
+ if (it == _policy_map.end()) {
Review Comment:
maybe remote is not s3, perhaps hdfs. check resource type is better.
##########
be/src/agent/task_worker_pool.cpp:
##########
@@ -437,6 +443,11 @@ void TaskWorkerPool::_drop_tablet_worker_thread_callback()
{
StorageEngine::instance()->txn_manager()->force_rollback_tablet_related_txns(
dropped_tablet->data_dir()->get_meta(),
drop_tablet_req.tablet_id,
drop_tablet_req.schema_hash, dropped_tablet->tablet_uid());
+ // We remove remote rowset directly.
+ // TODO(cyx): do remove in background
+ if (drop_tablet_req.is_drop_table_or_partition) {
+ dropped_tablet->remove_all_remote_rowsets();
Review Comment:
no return checker? how it will be when remove_all_remote_rowsets() is failed?
##########
be/src/agent/task_worker_pool.cpp:
##########
@@ -1683,9 +1701,64 @@ void
TaskWorkerPool::_submit_table_compaction_worker_thread_callback() {
}
}
-void TaskWorkerPool::_storage_medium_migrate_v2_worker_thread_callback() {
+void TaskWorkerPool::_storage_refresh_storage_policy_worker_thread_callback() {
+ while (_is_work) {
+ _is_doing_work = false;
+ // wait at most report_task_interval_seconds, or being notified
+ std::unique_lock<std::mutex> worker_thread_lock(_worker_thread_lock);
+ _worker_thread_condition_variable.wait_for(
+ worker_thread_lock,
+
std::chrono::seconds(config::storage_refresh_storage_policy_task_interval_seconds));
+ if (!_is_work) {
+ break;
+ }
+
+ if (_master_info.network_address.port == 0) {
+ // port == 0 means not received heartbeat yet
+ // sleep a short time and try again
+ LOG(INFO)
+ << "waiting to receive first heartbeat from frontend
before doing task report";
+ continue;
+ }
+
+ _is_doing_work = true;
+
+ TGetStoragePolicyResult result;
+ Status status = _master_client->refresh_storage_policy(&result);
+ if (!status.ok()) {
+ LOG(WARNING) << "refresh storage policy status not ok. ";
+ } else if (result.status.status_code != TStatusCode::OK) {
+ LOG(WARNING) << "refresh storage policy result status status_code
not ok. ";
+ } else {
+ // update storage policy mgr.
+ StoragePolicyMgr* spm =
ExecEnv::GetInstance()->storage_policy_mgr();
+ for (const auto& iter : result.result_entrys) {
+ shared_ptr<StoragePolicy> policy_ptr =
make_shared<StoragePolicy>();
+ policy_ptr->storage_policy_name = iter.policy_name;
+ policy_ptr->cooldown_datetime = iter.cooldown_datetime;
+ policy_ptr->cooldown_ttl = iter.cooldown_ttl;
+ policy_ptr->s3_endpoint = iter.s3_storage_param.s3_endpoint;
+ policy_ptr->s3_region = iter.s3_storage_param.s3_region;
+ policy_ptr->s3_ak = iter.s3_storage_param.s3_ak;
+ policy_ptr->s3_sk = iter.s3_storage_param.s3_sk;
+ policy_ptr->root_path = iter.s3_storage_param.root_path;
+ policy_ptr->bucket = iter.s3_storage_param.bucket;
+ policy_ptr->s3_conn_timeout_ms =
iter.s3_storage_param.s3_conn_timeout_ms;
+ policy_ptr->s3_max_conn = iter.s3_storage_param.s3_max_conn;
+ policy_ptr->s3_request_timeout_ms =
iter.s3_storage_param.s3_request_timeout_ms;
+ policy_ptr->md5_sum = iter.md5_checksum;
+
+ LOG(INFO) << "refresh storage policy task, policy " <<
*policy_ptr;
+ spm->periodic_put(iter.policy_name, std::move(policy_ptr));
+ }
+ }
+ }
+}
+
+void TaskWorkerPool::_storage_update_storage_policy_worker_thread_callback() {
Review Comment:
_storage_refresh_storage_policy_worker_thread_callback() and
_storage_update_storage_policy_worker_thread_callback() has too many duplicate
code, create a sub function is better.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: commits-unsubscr...@doris.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@doris.apache.org
For additional commands, e-mail: commits-h...@doris.apache.org
