diff --git a/docs/nightly/en/reference/sql/explain.md b/docs/nightly/en/reference/sql/explain.md index 82b545ab8..d0e384115 100644 --- a/docs/nightly/en/reference/sql/explain.md +++ b/docs/nightly/en/reference/sql/explain.md @@ -35,21 +35,42 @@ The column `plan_type` indicates whether it's a`logical_plan` or `physical_plan` Explains the execution of the plan by `ANALYZE`: ```sql -EXPLAIN ANALYZE SELECT * FROM monitor where host='host1'; +EXPLAIN ANALYZE SELECT * FROM monitor where host='host1'\G ``` ```sql -| plan_type | plan -| Plan with Metrics | CoalescePartitionsExec, metrics=[output_rows=1, elapsed_compute=79.167µs, spill_count=0, spilled_bytes=0, mem_used=0] - ProjectionExec: expr=[host@0 as host, ts@1 as ts, cpu@2 as cpu, memory@3 as memory], metrics=[output_rows=1, elapsed_compute=17.176µs, spill_count=0, spilled_bytes=0, mem_used=0] - CoalesceBatchesExec: target_batch_size=4096, metrics=[output_rows=1, elapsed_compute=14.248µs, spill_count=0, spilled_bytes=0, mem_used=0] - CoalesceBatchesExec: target_batch_size=4096, metrics=[output_rows=1, elapsed_compute=17.21µs, spill_count=0, spilled_bytes=0, mem_used=0] - FilterExec: host@0 = host1, metrics=[output_rows=1, elapsed_compute=541.801µs, spill_count=0, spilled_bytes=0, mem_used=0] - CoalesceBatchesExec: target_batch_size=4096, metrics=[output_rows=3, elapsed_compute=43.004µs, spill_count=0, spilled_bytes=0, mem_used=0] - RepartitionExec: partitioning=RoundRobinBatch(10), metrics=[fetch_time=5.832958ms, repart_time=10ns, send_time=2.467µs] - RepartitionExec: partitioning=RoundRobinBatch(10), metrics=[fetch_time=386.585µs, repart_time=1ns, send_time=7.833µs] - ExecutionPlan(PlaceHolder), metrics=[] - +*************************** 1. row *************************** +stage: 0 + node: 0 + plan: MergeScanExec: peers=[5471788335104(1274, 0), 5471788335105(1274, 1), ] metrics=[output_rows: 0, greptime_exec_read_cost: 0, finish_time: 1496211, ready_time: 846828, first_consume_time: 1491941, ] + +*************************** 2. row *************************** +stage: 1 + node: 0 + plan: CoalesceBatchesExec: target_batch_size=8192 metrics=[output_rows: 0, elapsed_compute: 4147, ] + FilterExec: host@2 = host1 metrics=[output_rows: 0, elapsed_compute: 32, ] + RepartitionExec: partitioning=RoundRobinBatch(32), input_partitions=8 metrics=[repart_time: 8, fetch_time: 230515, send_time: 256, ] + SeqScan: region=5471788335105(1274, 1), partition_count=0 (0 memtable ranges, 0 file ranges) metrics=[output_rows: 0, mem_used: 0, ] + +*************************** 3. row *************************** +stage: 1 + node: 1 + plan: CoalesceBatchesExec: target_batch_size=8192 metrics=[output_rows: 0, elapsed_compute: 3660, ] + FilterExec: host@2 = host1 metrics=[output_rows: 0, elapsed_compute: 32, ] + RepartitionExec: partitioning=RoundRobinBatch(32), input_partitions=8 metrics=[repart_time: 8, fetch_time: 113774, send_time: 256, ] + SeqScan: region=5471788335104(1274, 0), partition_count=0 (0 memtable ranges, 0 file ranges) metrics=[output_rows: 0, mem_used: 0, ] + +*************************** 4. row *************************** +stage: NULL + node: NULL + plan: Total rows: 0 +4 rows in set (0.002 sec) ``` -TODO: explains the output of `ANALYZE`. +In above example, we got 4 rows of output. Except the last row gives an overall summary, each remains row represents a staged node of the execution plan. The `stage` column indicates the stage number, and the `node` column indicates the node number. The `plan` column explains the plan and related metrics in detail. + +### Stage and node + +This example is based on a distributed partition table with two regions. We mark different "stage" in a distributed execution plan with stage number, start from 0. And in each stage, we may have multiple "node" to execute the plan in parallel. The node number also starts from 0 in each stage. These two words are only used here currently. Here is an example about 3 "node" with 2 "stage": + +![stage and node](/explain-stage-and-node.png) diff --git a/docs/nightly/zh/reference/sql/explain.md b/docs/nightly/zh/reference/sql/explain.md index 8ae708d30..9b48314d7 100644 --- a/docs/nightly/zh/reference/sql/explain.md +++ b/docs/nightly/zh/reference/sql/explain.md @@ -33,21 +33,42 @@ EXPLAIN SELECT * FROM monitor where host='host1'; 使用 `ANALYZE` 解释执行计划： ```sql -EXPLAIN ANALYZE SELECT * FROM monitor where host='host1'; +EXPLAIN ANALYZE SELECT * FROM monitor where host='host1'\G ``` ```sql -| plan_type | plan -| Plan with Metrics | CoalescePartitionsExec, metrics=[output_rows=1, elapsed_compute=79.167µs, spill_count=0, spilled_bytes=0, mem_used=0] - ProjectionExec: expr=[host@0 as host, ts@1 as ts, cpu@2 as cpu, memory@3 as memory], metrics=[output_rows=1, elapsed_compute=17.176µs, spill_count=0, spilled_bytes=0, mem_used=0] - CoalesceBatchesExec: target_batch_size=4096, metrics=[output_rows=1, elapsed_compute=14.248µs, spill_count=0, spilled_bytes=0, mem_used=0] - CoalesceBatchesExec: target_batch_size=4096, metrics=[output_rows=1, elapsed_compute=17.21µs, spill_count=0, spilled_bytes=0, mem_used=0] - FilterExec: host@0 = host1, metrics=[output_rows=1, elapsed_compute=541.801µs, spill_count=0, spilled_bytes=0, mem_used=0] - CoalesceBatchesExec: target_batch_size=4096, metrics=[output_rows=3, elapsed_compute=43.004µs, spill_count=0, spilled_bytes=0, mem_used=0] - RepartitionExec: partitioning=RoundRobinBatch(10), metrics=[fetch_time=5.832958ms, repart_time=10ns, send_time=2.467µs] - RepartitionExec: partitioning=RoundRobinBatch(10), metrics=[fetch_time=386.585µs, repart_time=1ns, send_time=7.833µs] - ExecutionPlan(PlaceHolder), metrics=[] - +*************************** 1. row *************************** +stage: 0 + node: 0 + plan: MergeScanExec: peers=[5471788335104(1274, 0), 5471788335105(1274, 1), ] metrics=[output_rows: 0, greptime_exec_read_cost: 0, finish_time: 1496211, ready_time: 846828, first_consume_time: 1491941, ] + +*************************** 2. row *************************** +stage: 1 + node: 0 + plan: CoalesceBatchesExec: target_batch_size=8192 metrics=[output_rows: 0, elapsed_compute: 4147, ] + FilterExec: host@2 = host1 metrics=[output_rows: 0, elapsed_compute: 32, ] + RepartitionExec: partitioning=RoundRobinBatch(32), input_partitions=8 metrics=[repart_time: 8, fetch_time: 230515, send_time: 256, ] + SeqScan: region=5471788335105(1274, 1), partition_count=0 (0 memtable ranges, 0 file ranges) metrics=[output_rows: 0, mem_used: 0, ] + +*************************** 3. row *************************** +stage: 1 + node: 1 + plan: CoalesceBatchesExec: target_batch_size=8192 metrics=[output_rows: 0, elapsed_compute: 3660, ] + FilterExec: host@2 = host1 metrics=[output_rows: 0, elapsed_compute: 32, ] + RepartitionExec: partitioning=RoundRobinBatch(32), input_partitions=8 metrics=[repart_time: 8, fetch_time: 113774, send_time: 256, ] + SeqScan: region=5471788335104(1274, 0), partition_count=0 (0 memtable ranges, 0 file ranges) metrics=[output_rows: 0, mem_used: 0, ] + +*************************** 4. row *************************** +stage: NULL + node: NULL + plan: Total rows: 0 +4 rows in set (0.002 sec) ``` - +在上面的例子中，我们得到了 4 行输出。除了最后一行给出了总体摘要，每一行都代表了执行计划的一个阶段节点。`stage` 列表示阶段编号，`node` 列表示节点编号。`plan` 列详细说明了计划和相关指标。 + +## Stage 和 node + +这个例子基于一个有两个 region 的分布式分区表。我们使用 stage 编号在分布式执行计划中标记不同的 stage，从 0 开始。在每个 stage 中，我们可能有多个 node 并行执行计划。node 编号在每个 stage 中也从 0 开始。以下是有 2 个 stage 和 3 个 node 的示例： + +![stage and node](/explain-stage-and-node.png)