Add a --shard-count argument to build-and-run.sh

The README suggests using a shard count that is divisible by the number
of nodes in your cluster. This is good advice, but we did not allow
changing the shardcount that we set by default.

build.sh

@@ -10,7 +10,7 @@ mkdir -p results/
 psql -v "ON_ERROR_STOP=1" -f sql/drop-tables.sql
 
 # set Citus configurations
-psql -c "ALTER ROLE current_user SET citus.shard_count TO 40" 2>/dev/null || true
+psql -c "ALTER ROLE current_user SET citus.shard_count TO $SHARD_COUNT" 2>/dev/null || true
 psql -c "ALTER ROLE current_user SET citus.enable_repartition_joins to on" 2>/dev/null || true
 
 # build hammerdb related tables

parse-arguments.sh

@@ -6,13 +6,41 @@ set -e
 # fail if a command that is piped fails
 set -o pipefail
 
-helpstring="usage: [--hammerdb-version[=]<version>] [--ch|--ch-queries-only] [--no-citus] [--name[=]name]"
+helpstring="Usage: [--hammerdb-version[=]<version>] [--ch|--ch-queries-only] [--no-citus] [--name[=]name] --shard-count=<shard_count>
+
+  --hammerdb-version    What version of HammerDB to use to run the benchmark.
+                        To be able to run Citus benchmarks it is required to
+                        use '4.3-custom' here. All other versions only work for
+                        standard postgres benchmarks.
+                        default: 4.3-custom
+  --ch                  Run both HammerDB TPROC-C and CH-benCHmark queries at
+                        the same time.
+  --ch-queries-only     Run only CH-benCHmark queries (so don't run HammerDB
+                        TPROC-C queries
+  --no-citus            Build the dataset and run the benchmark for standard
+                        Postgres instead of for Citus. This also requires
+                        setting pg_cituscompat to false in run.tcl
+  --name                The name you want to give to the benchmark run. This
+                        determines the filenames in the results directory.
+                        default: timestamp of start of benchmark
+  --shard-count         The amount of shards each distributed table should
+                        have. It's important to make sure that this is
+                        divisible by the number of worker nodes that you have,
+                        otherwise the load will not be evenly spread across
+                        nodes.
+                        default: 48
+"
 
 IS_CH_ONLY=${IS_CH_ONLY:-false}
 IS_CH_AND_TPCC=${IS_CH_AND_TPCC:-false}
 IS_TPCC=${IS_TPCC:-true}
 IS_CH=${IS_CH:-false}
 IS_CITUS=${IS_CITUS:-true}
+SHARD_COUNT=${SHARD_COUNT:-48}
+isodate=$(date +"%Y-%m-%dT%H:%M:%S")
+BENCHNAME=${BENCHNAME:-${isodate}}
+HAMMERDB_VERSION=${HAMMERDB_VERSION:-4.3-custom}
+SHARD_COUNT=${SHARD_COUNT:-48}
 
 # inspired by: https://stackoverflow.com/a/7680682/2570866
 optspec=":h-:"
@@ -25,8 +53,8 @@ while getopts "$optspec" optchar; do
                         echo "error: --hamerdb-version requires argument" >&2
                         exit 2
                     fi
-                    HAMMERDB_VERSION="${!OPTIND}"; OPTIND=$(( OPTIND + 1 ))
-                    BENCHNAME="$OPTARG"
+                    HAMMERDB_VERSION="${!OPTIND}"
+                    OPTIND=$(( OPTIND + 1 ))
                     ;;
                 hammerdb-version=*)
                     HAMMERDB_VERSION=${OPTARG#*=}
@@ -36,7 +64,19 @@ while getopts "$optspec" optchar; do
                         echo "error: --name requires argument" >&2
                         exit 2
                     fi
-                    BENCHNAME="${!OPTIND}"; OPTIND=$(( OPTIND + 1 ))
+                    BENCHNAME="${!OPTIND}"
+                    OPTIND=$(( OPTIND + 1 ))
+                    ;;
+                name=*)
+                    BENCHNAME=${OPTARG#*=}
+                    ;;
+                name)
+                    if [ -z ${!OPTIND+x} ]; then
+                        echo "error: --shard-count requires argument" >&2
+                        exit 2
+                    fi
+                    SHARD_COUNT="${!OPTIND}"
+                    OPTIND=$(( OPTIND + 1 ))
                     ;;
                 name=*)
                     BENCHNAME=${OPTARG#*=}
@@ -88,10 +128,6 @@ if [ -z "${PGPASSWORD+x}" ]; then
 fi
 export PGPASSWORD=${PGPASSWORD}
 
-isodate=$(date +"%Y-%m-%dT%H:%M:%S")
-BENCHNAME=${BENCHNAME:-${isodate}}
-HAMMERDB_VERSION=${HAMMERDB_VERSION:-4.3-custom}
-
 # Only parse the variables once and then export them to any nested scripts,
 # this way the BENCHNAME is the same for both the build and the run when
 # sourcing parse-arguments.sh from build-and-run.sh.