RONDB-205: Add options --mysqld-instrumentation and --extra-packages

Dockerfile

@@ -13,6 +13,8 @@ ARG TARGETPLATFORM
 ARG TARGETARCH
 ARG TARGETVARIANT
 
+ARG EXTRA_PACKAGES
+
 ARG OPEN_SSL_VERSION=1.1.1s
 
 RUN echo "Running on $BUILDPLATFORM, building for $TARGETPLATFORM"
@@ -22,6 +24,7 @@ RUN --mount=type=cache,target=/var/cache/apt,id=ubuntu22-apt \
     --mount=type=cache,target=/var/lib/apt/lists,id=ubuntu22-apt-lists \
     apt-get update -y \
     && apt-get install -y wget tar gzip \
+    $EXTRA_PACKAGES \
     libncurses5 libnuma-dev \
     bc
     # bc is required by dbt2

README.md

@@ -102,6 +102,27 @@ It may be the case that the benchmarks require more resources than are configure
 
 ***Note***: Benchmarking RonDB with a docker-compose setup on a single machine may not bring optimal performance results. This is because both the mysqlds and the ndbmtds (multi-threaded data nodes) scale in performance with more CPUs. In a production setting, each of these programs would be deployed on their own VM, whereby mysqlds and ndbmtds will scale linearly with up to 32 cores. The possibility of benchmarking was added here to give the user an introduction of benchmarking RonDB without needing to spin up a cluster with VMs.
 
+## Instrumentation for mysqld
+
+Instrumenting `mysqld` is recommended only for advanced debugging scenarios.
+The `--mysqld-instrumentation` argument is used to prepend the command line that starts `mysqld`.
+Use the `--extra-packages` argument to list the packages you need installed for this purpose.
+Example:
+
+```bash
+# Use valgrind to check for memory leaks during a benchmark run. Since we use the
+# --volumes-in-local-dir flag, the memcheck output will be saved under
+# `autogenerated_files/*/volumes/mysqlFilesDir_mysqld_*/`.
+./build_run_docker.sh \
+  --rondb-tarball-is-local \
+  -ruri ./rondb-21.04.9-linux-glibc2.35-arm64_v8.tar.gz \
+  -v 21.04.9 -m 1 -g 1 -r 2 -my 1 -a 1 \
+  --mysqld-instrumentation "valgrind --log-file=/srv/hops/mysql-cluster/mysql-files/memcheck.log" \
+  --extra-packages valgrind \
+  --volumes-in-local-dir \
+  --run-benchmark sysbench_single
+```
+
 ## Goals of this repository
 
 1. Create an image with RonDB installed "hopsworks/rondb-standalone:21.04.9"

build_run_docker.sh

@@ -41,6 +41,8 @@ Usage: $0
                                                     ]
     [-rtarl     --rondb-tarball-is-local            ]
     [-lv        --volumes-in-local-dir              ]
+    [-mi        --mysqld-instrumentation            ]
+    [-ep        --extra-packages                    ]
 EOF
 }
 
@@ -62,6 +64,8 @@ REPLICATION_FACTOR=1
 NODE_GROUPS=1
 RUN_BENCHMARK=
 VOLUME_TYPE=docker
+MYSQLD_INSTRUMENTATION=
+EXTRA_PACKAGES=
 
 POSITIONAL=()
 while [[ $# -gt 0 ]]; do
@@ -124,28 +128,45 @@ while [[ $# -gt 0 ]]; do
         shift # past argument
         ;;
 
+    -mi | --mysqld-instrumentation)
+        MYSQLD_INSTRUMENTATION="$2"
+        shift # past argument
+        shift # past value
+        ;;
+
+    -ep | --extra-packages)
+        EXTRA_PACKAGES="$2"
+        shift # past argument
+        shift # past value
+        ;;
+
     *)                     # unknown option
         POSITIONAL+=("$1") # save it in an array for later
         shift              # past argument
         ;;
     esac
 done
 
-echo "#################"
-echo "Parsed arguments:"
-echo "#################"
-echo
-echo "RonDB version                 = ${RONDB_VERSION}"
-echo "RonDB tarball local/remote    = ${RONDB_TARBALL_LOCAL_REMOTE}"
-echo "RonDB tarball URI             = ${RONDB_TARBALL_URI}"
-echo "Number of management nodes    = ${NUM_MGM_NODES}"
-echo "Node groups                   = ${NODE_GROUPS}"
-echo "Replication factor            = ${REPLICATION_FACTOR}"
-echo "Number of mysql nodes         = ${NUM_MYSQL_NODES}"
-echo "Number of api nodes           = ${NUM_API_NODES}"
-echo "Run benchmark                 = ${RUN_BENCHMARK}"
-echo "Volume type docker/local      = ${VOLUME_TYPE}"
-echo
+print-parsed-arguments(){
+    echo "#################"
+    echo "Parsed arguments:"
+    echo "#################"
+    echo
+    echo "RonDB version                 = ${RONDB_VERSION}"
+    echo "RonDB tarball local/remote    = ${RONDB_TARBALL_LOCAL_REMOTE}"
+    echo "RonDB tarball URI             = ${RONDB_TARBALL_URI}"
+    echo "Number of management nodes    = ${NUM_MGM_NODES}"
+    echo "Node groups                   = ${NODE_GROUPS}"
+    echo "Replication factor            = ${REPLICATION_FACTOR}"
+    echo "Number of mysql nodes         = ${NUM_MYSQL_NODES}"
+    echo "Number of api nodes           = ${NUM_API_NODES}"
+    echo "Run benchmark                 = ${RUN_BENCHMARK}"
+    echo "Volume type docker/local      = ${VOLUME_TYPE}"
+    echo "Instrumentation mysqld        = ${MYSQLD_INSTRUMENTATION}"
+    echo "Extra packages                = ${EXTRA_PACKAGES}"
+    echo
+}
+print-parsed-arguments
 
 set -- "${POSITIONAL[@]}" # restore unknown options
 if [[ -n $1 ]]; then
@@ -219,17 +240,29 @@ fi
 
 # We use this for the docker-compose project name, which will not allow "."
 RONDB_VERSION_NO_DOT=$(echo "$RONDB_VERSION" | tr -d '.')
+# We can't put an arbitrary command line or arbitrary list of packages as part
+# of a filename, so hash them. User can refer to parsed_arguments.txt for full
+# context.
+MYSQLD_INSTRUMENTATION_INDICATOR=
+if [ -n "$MYSQLD_INSTRUMENTATION" ]; then
+    MYSQLD_INSTRUMENTATION_INDICATOR="$(echo "$MYSQLD_INSTRUMENTATION" | sha256sum | sed -r 's/^(.{8}).*$/_mi-\1/')"
+fi
+EXTRA_PACKAGES_INDICATOR=
+if [ -n "$EXTRA_PACKAGES" ]; then
+    EXTRA_PACKAGES_INDICATOR="$(echo "$EXTRA_PACKAGES" | sha256sum | sed -r 's/^(.{8}).*$/_ep-\1/')"
+fi
 
 ## Uncomment this for quicker testing
 # yes | docker container prune
 # yes | docker volume prune
 
-FILE_SUFFIX="v${RONDB_VERSION_NO_DOT}_m${NUM_MGM_NODES}_g${NODE_GROUPS}_r${REPLICATION_FACTOR}_my${NUM_MYSQL_NODES}_api${NUM_API_NODES}"
+FILE_SUFFIX="v${RONDB_VERSION_NO_DOT}_m${NUM_MGM_NODES}_g${NODE_GROUPS}_r${REPLICATION_FACTOR}_my${NUM_MYSQL_NODES}_api${NUM_API_NODES}${MYSQLD_INSTRUMENTATION_INDICATOR}${EXTRA_PACKAGES_INDICATOR}"
 
 # https://stackoverflow.com/a/246128/9068781
 SCRIPT_DIR=$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" &>/dev/null && pwd)
 AUTOGENERATED_FILES_DIR="$SCRIPT_DIR/autogenerated_files/$FILE_SUFFIX"
 mkdir -p $AUTOGENERATED_FILES_DIR
+print-parsed-arguments > "$AUTOGENERATED_FILES_DIR/parsed_arguments.txt"
 
 DOCKER_COMPOSE_FILEPATH="$AUTOGENERATED_FILES_DIR/docker_compose.yml"
 CONFIG_INI_FILEPATH="$AUTOGENERATED_FILES_DIR/config.ini"
@@ -273,6 +306,7 @@ docker buildx build . \
     --tag $RONDB_IMAGE_NAME \
     --build-arg RONDB_VERSION=$RONDB_VERSION \
     --build-arg RONDB_TARBALL_LOCAL_REMOTE=$RONDB_TARBALL_LOCAL_REMOTE \
+    --build-arg "EXTRA_PACKAGES=$EXTRA_PACKAGES" \
     --build-arg RONDB_TARBALL_URI=$RONDB_TARBALL_URI
 
 #######################
@@ -534,6 +568,10 @@ if [ $NUM_MYSQL_NODES -gt 0 ]; then
             env_var=$(printf "$ENV_VAR_TEMPLATE" "MYSQL_SETUP_APP" "1")
             template+="$env_var"
         fi
+        if [ -n "$MYSQLD_INSTRUMENTATION" ]; then
+            env_var="$(printf "$ENV_VAR_TEMPLATE" "MYSQLD_INSTRUMENTATION" "$MYSQLD_INSTRUMENTATION")"
+            template+="$env_var"
+        fi
 
         BASE_DOCKER_COMPOSE_FILE+="$template"
 

resources/entrypoints/mysqld.sh

@@ -74,8 +74,8 @@ echo '[Entrypoint] Database initialized'
 export MYSQLD_PARENT_PID=$$
 if [ -z $MYSQL_SETUP_APP ]; then
     echo '[Entrypoint] Not setting up app here; going straight to execution of mysqld'
-    echo "[Entrypoint] Running: $@"
-    exec "$@"
+    echo "[Entrypoint] Running: $MYSQLD_INSTRUMENTATION $@"
+    exec $MYSQLD_INSTRUMENTATION "$@"
 fi
 
 echo '[Entrypoint] Executing mysqld as daemon with no networking allowed...'
@@ -172,5 +172,5 @@ mysqladmin -uroot --password=$MYSQL_ROOT_PASSWORD shutdown --socket="$SOCKET"
 echo "[Entrypoint] Successfully shut down MySQLd"
 
 echo '[Entrypoint] MySQL init process done. Ready for start up.'
-echo "[Entrypoint] Running: $@"
-exec "$@"
+echo "[Entrypoint] Running: $MYSQLD_INSTRUMENTATION $@"
+exec $MYSQLD_INSTRUMENTATION "$@"