Files
greptimedb/src/query/src/dist_plan/planner.rs
Lei, HUANG 02af2cb3cd fix: limit compaction input num 0.15 (#7117)
* fix/limit-compaction-input-num-0.15:
 **Add File Count Limit and Debug Logging in Compaction Process**

 - **`run.rs`**: Introduced a new method `num_files` in `FileGroup` to count files. This aids in managing file limits during compaction.
 - **`twcs.rs`**: Implemented an environment variable `TWCS_MAX_INPUT_FILE_NUM` to limit the number of input files during compaction. Added debug logging to track the maximum compaction file number and info logging to enforce the file limit. Enhanced logic to skip large files in append mode and adjusted the
 compaction process to respect the new file count limit.

Signed-off-by: Lei, HUANG <mrsatangel@gmail.com>

* fix/limit-compaction-input-num-0.15:
 **Enhancements in `twcs.rs`**

 - Introduced a default value for `max_input_file_num` with `DEFAULT_MAX_INPUT_FILE_NUM` constant set to 32.
 - Added error handling for environment variable `TWCS_MAX_INPUT_FILE_NUM` using `warn` to log unrecognized values.
 - Improved logging in `TwcsPicker` to include the current total input files when enforcing the max input file number limit.

Signed-off-by: Lei, HUANG <mrsatangel@gmail.com>

* fix: typo

Signed-off-by: Lei, HUANG <mrsatangel@gmail.com>

---------

Signed-off-by: Lei, HUANG <mrsatangel@gmail.com>
2025-10-20 20:28:14 +08:00

280 lines
10 KiB
Rust

// Copyright 2023 Greptime Team
//
// Licensed 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.
//! [ExtensionPlanner] implementation for distributed planner
use std::sync::Arc;
use async_trait::async_trait;
use catalog::CatalogManagerRef;
use common_catalog::consts::{DEFAULT_CATALOG_NAME, DEFAULT_SCHEMA_NAME};
use datafusion::common::Result;
use datafusion::datasource::DefaultTableSource;
use datafusion::execution::context::SessionState;
use datafusion::physical_plan::ExecutionPlan;
use datafusion::physical_planner::{ExtensionPlanner, PhysicalPlanner};
use datafusion_common::tree_node::{TreeNode, TreeNodeRecursion, TreeNodeVisitor};
use datafusion_common::{DataFusionError, TableReference};
use datafusion_expr::{LogicalPlan, UserDefinedLogicalNode};
use session::context::QueryContext;
use snafu::{OptionExt, ResultExt};
use store_api::storage::RegionId;
pub use table::metadata::TableType;
use table::table::adapter::DfTableProviderAdapter;
use table::table_name::TableName;
use crate::dist_plan::merge_scan::{MergeScanExec, MergeScanLogicalPlan};
use crate::dist_plan::merge_sort::MergeSortLogicalPlan;
use crate::error::{CatalogSnafu, TableNotFoundSnafu};
use crate::region_query::RegionQueryHandlerRef;
/// Planner for convert merge sort logical plan to physical plan
///
/// it is currently a fallback to sort, and doesn't change the execution plan:
/// `MergeSort(MergeScan) -> Sort(MergeScan) - to physical plan -> ...`
/// It should be applied after `DistExtensionPlanner`
///
/// (Later when actually impl this merge sort)
///
/// We should ensure the number of partition is not smaller than the number of region at present. Otherwise this would result in incorrect output.
pub struct MergeSortExtensionPlanner {}
#[async_trait]
impl ExtensionPlanner for MergeSortExtensionPlanner {
async fn plan_extension(
&self,
planner: &dyn PhysicalPlanner,
node: &dyn UserDefinedLogicalNode,
_logical_inputs: &[&LogicalPlan],
physical_inputs: &[Arc<dyn ExecutionPlan>],
session_state: &SessionState,
) -> Result<Option<Arc<dyn ExecutionPlan>>> {
if let Some(merge_sort) = node.as_any().downcast_ref::<MergeSortLogicalPlan>() {
if let LogicalPlan::Extension(ext) = &merge_sort.input.as_ref()
&& ext
.node
.as_any()
.downcast_ref::<MergeScanLogicalPlan>()
.is_some()
{
let merge_scan_exec = physical_inputs
.first()
.and_then(|p| p.as_any().downcast_ref::<MergeScanExec>())
.ok_or(DataFusionError::Internal(format!(
"Expect MergeSort's input is a MergeScanExec, found {:?}",
physical_inputs
)))?;
let partition_cnt = merge_scan_exec.partition_count();
let region_cnt = merge_scan_exec.region_count();
// if partition >= region, we know that every partition stream of merge scan is ordered
// and we only need to do a merge sort, otherwise fallback to quick sort
let can_merge_sort = partition_cnt >= region_cnt;
if can_merge_sort {
// TODO(discord9): use `SortPreservingMergeExec here`
}
// for now merge sort only exist in logical plan, and have the same effect as `Sort`
// doesn't change the execution plan, this will change in the future
let ret = planner
.create_physical_plan(&merge_sort.clone().into_sort(), session_state)
.await?;
Ok(Some(ret))
} else {
Ok(None)
}
} else {
Ok(None)
}
}
}
pub struct DistExtensionPlanner {
catalog_manager: CatalogManagerRef,
region_query_handler: RegionQueryHandlerRef,
}
impl DistExtensionPlanner {
pub fn new(
catalog_manager: CatalogManagerRef,
region_query_handler: RegionQueryHandlerRef,
) -> Self {
Self {
catalog_manager,
region_query_handler,
}
}
}
#[async_trait]
impl ExtensionPlanner for DistExtensionPlanner {
async fn plan_extension(
&self,
planner: &dyn PhysicalPlanner,
node: &dyn UserDefinedLogicalNode,
_logical_inputs: &[&LogicalPlan],
_physical_inputs: &[Arc<dyn ExecutionPlan>],
session_state: &SessionState,
) -> Result<Option<Arc<dyn ExecutionPlan>>> {
let Some(merge_scan) = node.as_any().downcast_ref::<MergeScanLogicalPlan>() else {
return Ok(None);
};
let input_plan = merge_scan.input();
let fallback = |logical_plan| async move {
let optimized_plan = self.optimize_input_logical_plan(session_state, logical_plan)?;
planner
.create_physical_plan(&optimized_plan, session_state)
.await
.map(Some)
};
if merge_scan.is_placeholder() {
// ignore placeholder
return fallback(input_plan).await;
}
let optimized_plan = input_plan;
let Some(table_name) = Self::extract_full_table_name(input_plan)? else {
// no relation found in input plan, going to execute them locally
return fallback(optimized_plan).await;
};
let Ok(regions) = self.get_regions(&table_name).await else {
// no peers found, going to execute them locally
return fallback(optimized_plan).await;
};
// TODO(ruihang): generate different execution plans for different variant merge operation
let schema = optimized_plan.schema().as_ref().into();
let query_ctx = session_state
.config()
.get_extension()
.unwrap_or_else(QueryContext::arc);
let merge_scan_plan = MergeScanExec::new(
session_state,
table_name,
regions,
input_plan.clone(),
&schema,
self.region_query_handler.clone(),
query_ctx,
session_state.config().target_partitions(),
merge_scan.partition_cols().to_vec(),
)?;
Ok(Some(Arc::new(merge_scan_plan) as _))
}
}
impl DistExtensionPlanner {
/// Extract fully resolved table name from logical plan
fn extract_full_table_name(plan: &LogicalPlan) -> Result<Option<TableName>> {
let mut extractor = TableNameExtractor::default();
let _ = plan.visit(&mut extractor)?;
Ok(extractor.table_name)
}
async fn get_regions(&self, table_name: &TableName) -> Result<Vec<RegionId>> {
let table = self
.catalog_manager
.table(
&table_name.catalog_name,
&table_name.schema_name,
&table_name.table_name,
None,
)
.await
.context(CatalogSnafu)?
.with_context(|| TableNotFoundSnafu {
table: table_name.to_string(),
})?;
Ok(table.table_info().region_ids())
}
/// Input logical plan is analyzed. Thus only call logical optimizer to optimize it.
fn optimize_input_logical_plan(
&self,
session_state: &SessionState,
plan: &LogicalPlan,
) -> Result<LogicalPlan> {
let state = session_state.clone();
state.optimizer().optimize(plan.clone(), &state, |_, _| {})
}
}
/// Visitor to extract table name from logical plan (TableScan node)
#[derive(Default)]
struct TableNameExtractor {
pub table_name: Option<TableName>,
}
impl TreeNodeVisitor<'_> for TableNameExtractor {
type Node = LogicalPlan;
fn f_down(&mut self, node: &Self::Node) -> Result<TreeNodeRecursion> {
match node {
LogicalPlan::TableScan(scan) => {
if let Some(source) = scan.source.as_any().downcast_ref::<DefaultTableSource>() {
if let Some(provider) = source
.table_provider
.as_any()
.downcast_ref::<DfTableProviderAdapter>()
{
if provider.table().table_type() == TableType::Base {
let info = provider.table().table_info();
self.table_name = Some(TableName::new(
info.catalog_name.clone(),
info.schema_name.clone(),
info.name.clone(),
));
}
return Ok(TreeNodeRecursion::Stop);
}
}
match &scan.table_name {
TableReference::Full {
catalog,
schema,
table,
} => {
self.table_name = Some(TableName::new(
catalog.to_string(),
schema.to_string(),
table.to_string(),
));
Ok(TreeNodeRecursion::Stop)
}
// TODO(ruihang): Maybe the following two cases should not be valid
TableReference::Partial { schema, table } => {
self.table_name = Some(TableName::new(
DEFAULT_CATALOG_NAME.to_string(),
schema.to_string(),
table.to_string(),
));
Ok(TreeNodeRecursion::Stop)
}
TableReference::Bare { table } => {
self.table_name = Some(TableName::new(
DEFAULT_CATALOG_NAME.to_string(),
DEFAULT_SCHEMA_NAME.to_string(),
table.to_string(),
));
Ok(TreeNodeRecursion::Stop)
}
}
}
_ => Ok(TreeNodeRecursion::Continue),
}
}
}