Merge branch 'main' of https://github.com/datafuselabs/databend into add_runtime_bloom_filter_for_merge_into

Author: JackTan25

.github/actions/benchmark_cloud/action.yml

@@ -41,14 +41,31 @@ runs:
         sudo apt-get update -yq
         sudo apt-get install -yq python3
 
+    - name: Prepare
+      working-directory: benchmark/clickbench
+      shell: bash
+      id: prepare
+      env:
+        BENDSQL_DSN: "databend://${{ inputs.cloud_user }}:${{ inputs.cloud_password }}@${{ inputs.cloud_gateway }}:443"
+      run: |
+        if [[ "${{ inputs.dataset }}" == "load" ]]; then
+          echo "CREATE DATABASE IF NOT EXISTS load_test_${{ inputs.run_id }};" | bendsql
+          echo "database=load_test_${{ inputs.run_id }}" >> $GITHUB_OUTPUT
+          echo "tries=1" >> $GITHUB_OUTPUT
+        else
+          echo "database=clickbench" >> $GITHUB_OUTPUT
+          echo "tries=3" >> $GITHUB_OUTPUT
+        fi
+
     - name: Run Benchmark
       working-directory: benchmark/clickbench
       env:
         BENCHMARK_ID: ${{ inputs.run_id }}
         BENCHMARK_DATASET: ${{ inputs.dataset }}
         BENCHMARK_SIZE: ${{ inputs.size }}
         BENCHMARK_VERSION: ${{ inputs.version }}
-        BENCHMARK_DATABASE: clickbench
+        BENCHMARK_DATABASE: ${{ steps.prepare.outputs.database }}
+        BENCHMARK_TRIES: ${{ steps.prepare.outputs.tries }}
         CLOUD_USER: ${{ inputs.cloud_user }}
         CLOUD_PASSWORD: ${{ inputs.cloud_password }}
         CLOUD_GATEWAY: ${{ inputs.cloud_gateway }}

benchmark/clickbench/benchmark_cloud.sh

@@ -8,6 +8,7 @@ BENCHMARK_SIZE=${BENCHMARK_SIZE:-Small}
 BENCHMARK_CACHE_SIZE=${BENCHMARK_CACHE_SIZE:-0}
 BENCHMARK_VERSION=${BENCHMARK_VERSION:-}
 BENCHMARK_DATABASE=${BENCHMARK_DATABASE:-default}
+BENCHMARK_TRIES=${BENCHMARK_TRIES:-3}
 
 if [[ -z "${BENCHMARK_VERSION}" ]]; then
     echo "Please set BENCHMARK_VERSION to run the benchmark."
@@ -76,16 +77,18 @@ bendsql --query="select * from system.settings where value != default;" -o table
 echo "Running queries..."
 
 # analyze table
-echo "Analyze table..."
-bendsql <"${BENCHMARK_DATASET}/analyze.sql"
+if [[ -f "${BENCHMARK_DATASET}/analyze.sql" ]]; then
+    echo "Analyze table..."
+    bendsql <"${BENCHMARK_DATASET}/analyze.sql"
+fi
 
 function run_query() {
     local query_num=$1
     local seq=$2
     local query=$3
 
     local q_time
-    q_time=$(echo "$query" | bendsql --time=server)
+    q_time=$(bendsql --time=server <"$query")
     if [[ -n $q_time ]]; then
         echo "Q${query_num}[$seq] succeeded in $q_time seconds"
         yq -i ".result[${query_num}] += [${q_time}]" -o json result.json
@@ -94,16 +97,20 @@ function run_query() {
     fi
 }
 
-TRIES=3
 QUERY_NUM=0
-while read -r query; do
+for query in "${BENCHMARK_DATASET}"/queries/*.sql; do
     echo "Running Q${QUERY_NUM}: ${query}"
     yq -i ".result += [[]]" -o json result.json
-    for i in $(seq 1 $TRIES); do
+    for i in $(seq 1 "$BENCHMARK_TRIES"); do
         run_query "$QUERY_NUM" "$i" "$query"
     done
     QUERY_NUM=$((QUERY_NUM + 1))
-done <"${BENCHMARK_DATASET}/queries.sql"
+done
 
 echo "Cleaning up..."
+if [[ "${BENCHMARK_DATASET}" == "load" ]]; then
+    echo "Dropping database..."
+    echo "DROP DATABASE IF EXISTS ${BENCHMARK_DATABASE};" | bendsql
+fi
+echo "Drop warehouse..."
 echo "DROP WAREHOUSE IF EXISTS '${CLOUD_WAREHOUSE}';" | bendsql

benchmark/clickbench/benchmark_local.sh

@@ -70,17 +70,16 @@ instance_type=$(curl -H "X-aws-ec2-metadata-token: $token" http://169.254.169.25
 echo "Instance type: ${instance_type}"
 
 # Load data
-if [ "${BENCHMARK_DATASET}" == "internal" ]; then
+echo "Loading data..."
+load_start=$(date +%s)
+if [[ -f "${BENCHMARK_DATASET}/load.sh" ]]; then
     bash "${BENCHMARK_DATASET}"/load.sh
-    load_time=0
 else
-    echo "Loading data..."
-    load_start=$(date +%s)
     bendsql <"${BENCHMARK_DATASET}/load.sql"
-    load_end=$(date +%s)
-    load_time=$(python3 -c "print($load_end - $load_start)")
-    echo "Data loaded in ${load_time}s."
 fi
+load_end=$(date +%s)
+load_time=$(python3 -c "print($load_end - $load_start)")
+echo "Data loaded in ${load_time}s."
 
 data_size=$(echo "select sum(data_compressed_size) from system.tables where database = '${BENCHMARK_DATASET}';" | bendsql -o tsv)
 
@@ -99,7 +98,7 @@ function run_query() {
     local query=$3
 
     local q_time
-    q_time=$(echo "$query" | bendsql --time=server)
+    q_time=$(bendsql --time=server <"$query")
     if [[ -n $q_time ]]; then
         echo "Q${query_num}[$seq] succeeded in $q_time seconds"
         yq -i ".result[${query_num}] += [${q_time}]" -o json result.json
@@ -110,7 +109,7 @@ function run_query() {
 
 TRIES=3
 QUERY_NUM=0
-while read -r query; do
+for query in "${BENCHMARK_DATASET}"/queries/*.sql; do
     echo "Running Q${QUERY_NUM}: ${query}"
     sync
     echo 3 | sudo tee /proc/sys/vm/drop_caches
@@ -119,4 +118,4 @@ while read -r query; do
         run_query "$QUERY_NUM" "$i" "$query"
     done
     QUERY_NUM=$((QUERY_NUM + 1))
-done <"${BENCHMARK_DATASET}/queries.sql"
+done

benchmark/clickbench/hits/queries.sql

@@ -0,0 +1,2 @@
+SELECT COUNT(*)
+FROM hits;

benchmark/clickbench/hits/queries/00.sql

@@ -0,0 +1,3 @@
+SELECT COUNT(*)
+FROM hits
+WHERE AdvEngineID <> 0;

benchmark/clickbench/hits/queries/01.sql

@@ -0,0 +1,4 @@
+SELECT SUM(AdvEngineID),
+    COUNT(*),
+    AVG(ResolutionWidth)
+FROM hits;

benchmark/clickbench/hits/queries/02.sql

@@ -0,0 +1,2 @@
+SELECT AVG(UserID)
+FROM hits;

benchmark/clickbench/hits/queries/03.sql

@@ -0,0 +1,2 @@
+SELECT COUNT(DISTINCT UserID)
+FROM hits;

benchmark/clickbench/hits/queries/04.sql

@@ -0,0 +1,2 @@
+SELECT COUNT(DISTINCT SearchPhrase)
+FROM hits;