hcq/oceanbase
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity
Files
ad7be2769266f9b0d9425121aa9b3230909bfb0b
oceanbase/src/sql/executor/ob_distributed_task_executor.cpp
oceanbase-admin cea7de1475 init push
2021-05-31 22:56:52 +08:00

378 lines
16 KiB
C++
Raw Blame History

/**
* Copyright (c) 2021 OceanBase
* OceanBase CE is licensed under Mulan PubL v2.
* You can use this software according to the terms and conditions of the Mulan PubL v2.
* You may obtain a copy of Mulan PubL v2 at:
* http://license.coscl.org.cn/MulanPubL-2.0
* THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND,
* EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT,
* MERCHANTABILITY OR FIT FOR A PARTICULAR PURPOSE.
* See the Mulan PubL v2 for more details.
*/
#define USING_LOG_PREFIX SQL_EXE
#include "sql/session/ob_sql_session_info.h"
#include "sql/executor/ob_distributed_task_executor.h"
#include "sql/executor/ob_distributed_transmit.h"
#include "sql/executor/ob_receive.h"
#include "sql/executor/ob_job.h"
#include "sql/executor/ob_task_executor_ctx.h"
#include "sql/executor/ob_executor_rpc_impl.h"
#include "sql/executor/ob_trans_result_collector.h"
#include "sql/engine/ob_physical_plan_ctx.h"
#include "sql/engine/ob_phy_operator.h"
#include "sql/engine/ob_exec_context.h"
#include "sql/engine/ob_des_exec_context.h"
#include "lib/utility/ob_tracepoint.h"
#include "sql/executor/ob_bkgd_dist_task.h"
#include "sql/executor/ob_executor_rpc_processor.h"
#include "rootserver/ob_root_service.h"
using namespace oceanbase::common;
namespace oceanbase {
namespace sql {
ObDistributedTaskExecutor::ObDistributedTaskExecutor(const uint64_t scheduler_id)
: scheduler_id_(scheduler_id), trans_result_(NULL)
{}
ObDistributedTaskExecutor::~ObDistributedTaskExecutor()
{}
int ObDistributedTaskExecutor::execute(ObExecContext& query_ctx, ObJob* job, ObTaskInfo* task_info)
{
int ret = OB_SUCCESS;
ObPhyOperator* root_op = NULL;
ObDistributedTransmitInput* trans_input = NULL;
ObIPhyOperatorInput* op_input = NULL;
ObExecutorRpcImpl* rpc = NULL;
ObExecContext* exec_ctx_snap = NULL;
ObSQLSessionInfo* session_snap = NULL;
ObPhysicalPlanCtx* plan_ctx_snap = NULL;
ObTask task;
if (OB_ISNULL(exec_ctx_snap = query_ctx.get_scheduler_thread_ctx().get_dis_exec_ctx_for_update())) {
ret = OB_ERR_UNEXPECTED;
LOG_ERROR("exec ctx snap is NULL", K(ret));
} else if (OB_ISNULL(session_snap = exec_ctx_snap->get_my_session()) ||
OB_ISNULL(plan_ctx_snap = exec_ctx_snap->get_physical_plan_ctx())) {
ret = OB_ERR_UNEXPECTED;
LOG_ERROR("session snap or plan ctx snap is NULL", K(ret), K(session_snap), K(plan_ctx_snap));
} else if (OB_I(t1)(OB_ISNULL(job) || OB_ISNULL(task_info))) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("job or taskinfo is not set", K(ret), K(job), K(task_info));
} else if (OB_I(t2)(OB_ISNULL(root_op = job->get_root_op()))) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("fail execute task. no root op found", K(ret), K(root_op));
} else if (OB_I(t3)(OB_UNLIKELY(!IS_DIST_TRANSMIT(root_op->get_type())))) {
ret = OB_ERR_UNEXPECTED;
LOG_ERROR("valid check fail. root op type must be ObTransmit", K(ret), K(root_op->get_type()));
} else if (OB_I(t4)(OB_ISNULL(op_input = exec_ctx_snap->get_phy_op_input(root_op->get_id())))) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("fail to get op input", K(ret), "op_id", root_op->get_id());
} else if (OB_UNLIKELY(!IS_DIST_TRANSMIT(op_input->get_phy_op_type()))) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("Optimizer should generate PHY_DISTRIBUTED_TRANSMIT for this plan",
K(ret),
"input_type",
op_input->get_phy_op_type());
} else if (OB_ISNULL(trans_input = static_cast<ObDistributedTransmitInput*>(op_input))) {
ret = OB_ERR_UNEXPECTED; // should never reach here
LOG_WARN("fail cast op", K(ret), "trans_input", trans_input);
} else if (OB_FAIL(OB_I(t6) ObTaskExecutorCtxUtil::get_task_executor_rpc(query_ctx, rpc))) {
LOG_WARN("fail get rpc", K(ret));
} else if (OB_ISNULL(rpc)) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("rpc is NULL", K(ret));
} else if (OB_FAIL(build_task(*exec_ctx_snap, *job, *task_info, task))) {
LOG_WARN("fail to build task", K(ret), K(*job), K(*task_info));
} else {
ObExecutorRpcCtx rpc_ctx(session_snap->get_rpc_tenant_id(),
plan_ctx_snap->get_timeout_timestamp(),
exec_ctx_snap->get_task_exec_ctx().get_min_cluster_version(),
&query_ctx.get_scheduler_thread_ctx().get_scheduler_retry_info_for_update(),
query_ctx.get_my_session(),
query_ctx.get_scheduler_thread_ctx().is_plain_select_stmt());
task_info->set_task_send_begin(ObTimeUtility::current_time());
task_info->set_state(OB_TASK_STATE_RUNNING);
trans_input->set_ob_task_id(task_info->get_task_location().get_ob_task_id());
if (OB_FAIL(task_dispatch(*exec_ctx_snap, *rpc, rpc_ctx, task, *task_info))) {
bool skip_failed_tasks = false;
int check_ret = OB_SUCCESS;
if (OB_SUCCESS != (check_ret = should_skip_failed_tasks(*task_info, skip_failed_tasks))) {
// check fail, set ret to check_ret
LOG_WARN("fail to check if should skip failed tasks", K(ret), K(check_ret), K(*job), K(rpc_ctx));
ret = check_ret;
} else if (true == skip_failed_tasks) {
// should skip failed tasks, log user warning and skip it, than return OB_ERR_TASK_SKIPPED
LOG_WARN("fail to do task on some server, log user warning and skip it",
K(ret),
K(task_info->get_task_location().get_server()),
K(*job),
K(rpc_ctx));
LOG_USER_WARN(OB_ERR_TASK_SKIPPED,
to_cstring(task_info->get_task_location().get_server()),
common::ob_errpkt_errno(ret, lib::is_oracle_mode()));
ret = OB_ERR_TASK_SKIPPED;
} else {
// let user see this ret
LOG_WARN("fail to submit task", K(ret), K(*task_info), K(rpc_ctx));
}
} else {
}
NG_TRACE_EXT(distributed_task_submited,
OB_ID(ret),
ret,
OB_ID(runner_svr),
task_info->get_task_location().get_server(),
OB_ID(task),
task);
}
return ret;
}
int ObDistributedTaskExecutor::kill(ObExecContext& query_ctx, ObJob* job, ObTaskInfo* task_info)
{
int ret = OB_SUCCESS;
ObPhyOperator* root_op = NULL;
ObExecutorRpcImpl* rpc = NULL;
const ObExecContext* exec_ctx_snap = NULL;
ObSQLSessionInfo* session_snap = NULL;
ObPhysicalPlanCtx* plan_ctx_snap = NULL;
if (OB_ISNULL(exec_ctx_snap = query_ctx.get_scheduler_thread_ctx().get_dis_exec_ctx())) {
ret = OB_ERR_UNEXPECTED;
LOG_ERROR("exec ctx snap is NULL", K(ret));
} else if (OB_ISNULL(session_snap = exec_ctx_snap->get_my_session()) ||
OB_ISNULL(plan_ctx_snap = exec_ctx_snap->get_physical_plan_ctx())) {
ret = OB_ERR_UNEXPECTED;
LOG_ERROR("session snap or plan ctx snap is NULL", K(ret), K(session_snap), K(plan_ctx_snap));
} else if (OB_I(t1)(OB_ISNULL(task_info) || OB_ISNULL(job))) {
ret = OB_INVALID_ARGUMENT;
LOG_WARN("param is NULL", K(ret), K(task_info), K(job));
} else if (OB_I(t2)(OB_ISNULL(root_op = job->get_root_op()))) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("fail execute task. no root op found", K(ret), K(*job));
} else if (OB_I(t3) OB_UNLIKELY(!IS_TRANSMIT(root_op->get_type()))) {
ret = OB_ERR_UNEXPECTED;
LOG_ERROR("valid check fail. root op type must be ObTransmit", K(ret), K(root_op->get_type()));
} else if (OB_FAIL(OB_I(t5) ObTaskExecutorCtxUtil::get_task_executor_rpc(query_ctx, rpc))) {
LOG_WARN("fail get rpc", K(ret));
} else if (OB_ISNULL(rpc)) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("rpc is NULL", K(ret));
} else {
ObExecutorRpcCtx rpc_ctx(session_snap->get_rpc_tenant_id(),
plan_ctx_snap->get_timeout_timestamp(),
exec_ctx_snap->get_task_exec_ctx().get_min_cluster_version(),
&query_ctx.get_scheduler_thread_ctx().get_scheduler_retry_info_for_update(),
query_ctx.get_my_session(),
query_ctx.get_scheduler_thread_ctx().is_plain_select_stmt());
if (OB_FAIL(rpc->task_kill(
rpc_ctx, task_info->get_task_location().get_ob_task_id(), task_info->get_task_location().get_server()))) {
LOG_WARN("fail to kill task", K(ret), K(*task_info), K(rpc_ctx));
} else {
}
}
return ret;
}
int ObDistributedTaskExecutor::close_result(ObExecContext& ctx, const ObTaskInfo* task_info)
{
int ret = OB_SUCCESS;
if (OB_ISNULL(task_info)) {
ret = OB_ERR_UNEXPECTED;
LOG_ERROR("task info is NULL", K(ret));
} else {
const ObIArray<ObSliceEvent>& slice_events = task_info->get_slice_events();
if (OB_UNLIKELY(slice_events.count() < 0)) {
ret = OB_ERR_UNEXPECTED;
LOG_ERROR("slice events count < 0", K(ret), K(slice_events.count()));
} else {
// const ObSliceEvent &slice_event = slice_events.at(0);
// if (OB_ISNULL(slice_event)) {
// ret = OB_ERR_UNEXPECTED;
// LOG_ERROR("slice event is NULL", K(ret), K(slice_events.count()));
// } else {
// The result has been pulled back to the local situation,
// and the remote end has released the result,
// so there is no need to send the rpc to release the result;
// if the results are not pulled back to the local situation,
// the remote end still saves the results, so rpc must be sent to release the remote results
// const ObTaskSmallResult sr = slice_event->get_small_result();
// if (!sr.has_data()) {
if (OB_FAIL(send_close_result_rpc(ctx, task_info))) {
LOG_WARN("fail to send close result rpc", K(ret), K(*task_info));
}
//}
// }
}
}
return ret;
}
int ObDistributedTaskExecutor::send_close_result_rpc(ObExecContext& ctx, const ObTaskInfo* task_info)
{
int ret = OB_SUCCESS;
if (OB_ISNULL(task_info)) {
ret = OB_ERR_UNEXPECTED;
LOG_ERROR("task info is NULL", K(ret));
} else {
const ObExecContext* exec_ctx_snap = NULL;
ObSQLSessionInfo* session_snap = NULL;
ObPhysicalPlanCtx* plan_ctx_snap = NULL;
ObExecutorRpcImpl* rpc = NULL;
const ObTaskLocation& task_loc = task_info->get_task_location();
ObSliceID ob_slice_id;
int bak_ret = OB_SUCCESS;
// When schedule error or early terminate (statement with limit clause), result events
// are not processed (still in %response_task_events_ queue), there is no slice event in task.
// In this case, we remove intermediate result with slice_id 0.
const ObIArray<ObSliceEvent>& slices = task_info->get_slice_events();
for (int64_t i = 0; i < std::max(slices.count(), 1L); i++) {
ret = OB_SUCCESS;
ob_slice_id.set_ob_task_id(task_loc.get_ob_task_id());
ob_slice_id.set_slice_id(slices.empty() ? 0 : slices.at(i).get_ob_slice_id().get_slice_id());
if (OB_ISNULL(exec_ctx_snap = ctx.get_scheduler_thread_ctx().get_dis_exec_ctx())) {
ret = OB_ERR_UNEXPECTED;
LOG_ERROR("exec ctx snap is NULL", K(ret));
} else if (OB_ISNULL(session_snap = exec_ctx_snap->get_my_session()) ||
OB_ISNULL(plan_ctx_snap = exec_ctx_snap->get_physical_plan_ctx())) {
ret = OB_ERR_UNEXPECTED;
LOG_ERROR("session snap or plan ctx snap is NULL", K(ret), K(session_snap), K(plan_ctx_snap));
} else if (OB_FAIL(ObTaskExecutorCtxUtil::get_task_executor_rpc(ctx, rpc))) {
LOG_ERROR("fail get rpc", K(ret));
} else if (OB_ISNULL(rpc)) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("rpc is NULL", K(ret));
} else {
ObExecutorRpcCtx rpc_ctx(session_snap->get_rpc_tenant_id(),
plan_ctx_snap->get_timeout_timestamp(),
exec_ctx_snap->get_task_exec_ctx().get_min_cluster_version(),
&ctx.get_scheduler_thread_ctx().get_scheduler_retry_info_for_update(),
ctx.get_my_session(),
ctx.get_scheduler_thread_ctx().is_plain_select_stmt());
if (OB_FAIL(rpc->close_result(rpc_ctx, ob_slice_id, task_loc.get_server()))) {
LOG_WARN("fail to rpc call close_result", K(ret), K(ob_slice_id), K(task_loc), K(rpc_ctx));
}
}
if (OB_FAIL(ret)) {
bak_ret = ret;
}
}
ret = bak_ret;
}
return ret;
}
int ObDistributedTaskExecutor::build_task(ObExecContext& query_ctx, ObJob& job, ObTaskInfo& task_info, ObTask& task)
{
int ret = OB_SUCCESS;
ObPhyOperator* root_op = NULL;
const ObPhysicalPlan* phy_plan = NULL;
if (OB_ISNULL(root_op = job.get_root_op())) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("root op is NULL", K(ret), K(job), K(task_info));
} else if (OB_ISNULL(phy_plan = root_op->get_phy_plan())) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("phy plan is NULL", K(ret), K(job), K(task_info));
} else if (OB_FAIL(OB_I(t1) build_task_op_input(query_ctx, task_info, *root_op))) {
LOG_WARN("fail to build op input", K(ret), K(task_info));
} else if (OB_ISNULL(query_ctx.get_physical_plan_ctx())) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("physical plan ctx is null", K(ret));
} else {
query_ctx.get_physical_plan_ctx()->set_phy_plan(phy_plan);
const ObTaskInfo::ObRangeLocation& range_loc = task_info.get_range_location();
for (int64_t i = 0; OB_SUCC(ret) && i < range_loc.part_locs_.count(); ++i) {
const ObTaskInfo::ObPartLoc& part_loc = range_loc.part_locs_.at(i);
if (OB_FAIL(task.add_partition_key(part_loc.partition_key_))) {
LOG_WARN("fail to add partition key into ObTask", K(ret), K(i), K(part_loc.partition_key_));
} else if (OB_FAIL(task.assign_ranges(part_loc.scan_ranges_))) {
LOG_WARN("assign range failed", K(ret));
} else {
// Add the partition key of the right table of mv
// so that it can be passed to the start participant interface at the remote end
for (int64_t j = 0; OB_SUCC(ret) && j < part_loc.depend_table_keys_.count(); ++j) {
if (OB_FAIL(task.add_partition_key(part_loc.depend_table_keys_.at(j)))) {
LOG_WARN("fail to add partition key into ObTask", K(ret), K(j), K(part_loc.depend_table_keys_.at(j)));
}
}
}
}
if (OB_SUCC(ret)) {
LOG_DEBUG("build task",
"ctrl_svr",
job.get_ob_job_id().get_server(),
"ob_task_id",
task_info.get_task_location().get_ob_task_id(),
"runner_svr",
task_info.get_task_location().get_server(),
"range_loc",
task_info.get_range_location());
task.set_ctrl_server(job.get_ob_job_id().get_server());
task.set_ob_task_id(task_info.get_task_location().get_ob_task_id());
task.set_location_idx(task_info.get_location_idx());
task.set_runner_server(task_info.get_task_location().get_server());
task.set_serialize_param(query_ctx, *root_op, *phy_plan);
}
}
return ret;
}
int ObDistributedTaskExecutor::task_dispatch(
ObExecContext& exec_ctx, ObExecutorRpcImpl& rpc, ObExecutorRpcCtx& rpc_ctx, ObTask& task, ObTaskInfo& task_info)
{
int ret = OB_SUCCESS;
if (!task_info.is_background()) {
if (OB_ISNULL(trans_result_)) {
ret = OB_NOT_INIT;
LOG_WARN("trans result is NULL", K(ret));
} else if (OB_FAIL(trans_result_->send_task(task_info))) {
LOG_WARN("send task failed", K(ret), K(task_info));
} else if (OB_FAIL(rpc.task_submit(
rpc_ctx, task, task_info.get_task_location().get_server(), trans_result_->get_trans_result()))) {
LOG_WARN("task submit failed", K(ret));
}
} else {
const int64_t size = task.get_serialize_size();
int64_t pos = 0;
char* buf = NULL;
if (OB_ISNULL(exec_ctx.get_my_session())) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("session is NULL", K(ret));
} else if (OB_ISNULL(buf = static_cast<char*>(exec_ctx.get_allocator().alloc(size)))) {
ret = OB_ALLOCATE_MEMORY_FAILED;
LOG_WARN("allocate memory failed", K(ret));
} else if (OB_FAIL(task.serialize(buf, size, pos))) {
LOG_WARN("task serialize failed", K(ret), K(task));
} else {
ObString serialized_task(static_cast<int32_t>(size), buf);
ObSchedBKGDDistTask sched_task;
if (OB_FAIL(sched_task.init(exec_ctx.get_my_session()->get_effective_tenant_id(),
rpc_ctx.get_timeout_timestamp(),
task.get_ob_task_id(),
scheduler_id_,
task_info.get_range_location().part_locs_.at(0).partition_key_,
task_info.get_task_location().get_server(),
serialized_task))) {
LOG_WARN("init task failed", K(ret));
} else if (OB_ISNULL(GCTX.root_service_)) {
ret = OB_ERR_UNEXPECTED;
LOG_WARN("root service is NULL", K(ret));
} else if (OB_FAIL(GCTX.root_service_->schedule_sql_bkgd_task(sched_task))) {
LOG_WARN("schedule background task failed", K(ret), K(sched_task));
} else {
LOG_INFO("start schedule background task", K(sched_task));
}
}
if (NULL != buf) {
exec_ctx.get_allocator().free(buf);
}
}
return ret;
}
} // namespace sql
} // namespace oceanbase
Reference in New Issue View Git Blame Copy Permalink