feat(upgrade): add forward version check for upgrade script

https://emqx.atlassian.net/browse/EMQX-6978

[Related EIP](c4864eeccb/active/0022-forward-check-install-upgrade-script.md)

Currently, when performing a hot upgrade, the scripts that are run are
those from the currently installed EMQX version. It has a validation
that prevents upgrade between different minor versions. If we want to
allow an upgrade from, say, 5.0.x to 5.1.y, then the scripts in
already released packages will deny such operation. Also, if an
upgrade installation script contains a bug in the current, it will
never be able to execute properly without manual patching.

By attempting to execute the scripts from the target version, we may
add fixes and new validations to new EMQX versions and have them
executed by older versions.

bin/emqx

@@ -5,7 +5,13 @@
 set -euo pipefail
 
 DEBUG="${DEBUG:-0}"
-[ "$DEBUG" -eq 1 ] && set -x
+if [ "$DEBUG" -eq 1 ]; then
+  set -x
+fi
+if [ "$DEBUG" -eq 2 ]; then
+  set -x
+  export PS4='+(${BASH_SOURCE}:${LINENO}): ${FUNCNAME[0]:+${FUNCNAME[0]}(): }'
+fi
 
 # We need to find real directory with emqx files on all platforms
 # even when bin/emqx is symlinked on several levels
@@ -36,6 +42,7 @@ export RUNNER_ROOT_DIR
 export EMQX_ETC_DIR
 export REL_VSN
 export SCHEMA_MOD
+export IS_ENTERPRISE
 
 RUNNER_SCRIPT="$RUNNER_BIN_DIR/$REL_NAME"
 CODE_LOADING_MODE="${CODE_LOADING_MODE:-embedded}"
@@ -540,6 +547,121 @@ check_license() {
     fi
 }
 
+# When deciding which install upgrade script to run, we have to check
+# our own version so we may avoid infinite loops and call the correct
+# version.
+current_script_version() {
+  curr_script=$(basename "${BASH_SOURCE[0]}")
+  suffix=${curr_script#*-}
+  if [[ "${suffix}" == "${curr_script}" ]]; then
+    # there's no suffix, so we're running the default `emqx` script;
+    # we'll have to trust the REL_VSN variable
+    echo "$REL_VSN"
+  else
+    echo "${suffix}"
+  fi
+}
+
+parse_semver() {
+    echo "$1" | tr '.|-' ' '
+}
+
+max_version_of() {
+  local vsn1="$1"
+  local vsn2="$2"
+
+  echo "${vsn1}" "${vsn2}" | tr " " "\n" | sort -rV | head -n1
+}
+
+versioned_script_path() {
+  local script_name="$1"
+  local vsn="$2"
+
+  echo "$RUNNER_ROOT_DIR/bin/$script_name-$vsn"
+}
+
+does_script_version_exist() {
+  local script_name="$1"
+  local vsn="$2"
+
+  if [[ -f "$(versioned_script_path "$script_name" "$vsn")" ]]; then
+    return 0
+  else
+    return 1
+  fi
+}
+
+# extract_from_package packege_path destination file1 file2
+extract_from_package() {
+  local package="$1"
+  local dest_dir="$2"
+  shift 2
+
+  tar -C "$dest_dir" -xf "$package" "$@"
+}
+
+am_i_the_newest_script() {
+  local curr_vsn other_vsn
+  curr_vsn="$(current_script_version)"
+  other_vsn="$1"
+  max_vsn="$(max_version_of "$other_vsn" "$curr_vsn")"
+
+  if [[ "$max_vsn" == "$curr_vsn" ]]; then
+    return 0
+  else
+    return 1
+  fi
+}
+
+locate_package() {
+  local package_path candidates vsn
+  vsn="$1"
+
+  if [[ "${IS_ENTERPRISE}" == "yes" ]]; then
+    package_pattern="$RUNNER_ROOT_DIR/releases/emqx-enterprise-$vsn-*.tar.gz"
+  else
+    package_pattern="$RUNNER_ROOT_DIR/releases/emqx-$vsn-*.tar.gz"
+  fi
+
+  # shellcheck disable=SC2207,SC2086
+  candidates=($(ls $package_pattern))
+
+  if [[ "${#candidates[@]}" == 0 ]]; then
+    logerr "No package matching $package_pattern found."
+    exit 1
+  elif [[ "${#candidates[@]}" -gt 1 ]]; then
+    logerr "Multiple packages matching $package_pattern found.  Ensure only one exists."
+    exit 1
+  else
+    echo "${candidates[0]}"
+  fi
+}
+
+ensure_newest_script_is_extracted() {
+  local newest_vsn="$1"
+  local package_path tmpdir
+
+  if does_script_version_exist "emqx" "$newest_vsn" \
+     && does_script_version_exist "install_upgrade.escript" "$newest_vsn"; then
+    return
+  else
+    package_path="$(locate_package "$newest_vsn")"
+    tmpdir="$(mktemp -dp /tmp emqx.XXXXXXXXXXX)"
+
+    extract_from_package \
+      "$package_path" \
+      "$tmpdir" \
+      "bin/emqx-$newest_vsn" \
+      "bin/install_upgrade.escript-$newest_vsn"
+
+    cp "$tmpdir/bin/emqx-$newest_vsn" \
+       "$tmpdir/bin/install_upgrade.escript-$newest_vsn" \
+       "$RUNNER_ROOT_DIR/bin/"
+
+    rm -rf "$tmpdir"
+  fi
+}
+
 # Run an escript in the node's environment
 relx_escript() {
     shift; scriptpath="$1"; shift
@@ -922,8 +1044,20 @@ case "${COMMAND}" in
 
         assert_node_alive
 
+        curr_vsn="$(current_script_version)"
+        target_vsn="$1"
+        newest_vsn="$(max_version_of "$target_vsn" "$curr_vsn")"
+        ensure_newest_script_is_extracted "$newest_vsn"
+        # if we are not the newest script, run the same command from it
+        if ! am_i_the_newest_script "$newest_vsn"; then
+          script_path="$(versioned_script_path emqx "$newest_vsn")"
+          exec "$script_path" "$COMMAND" "$@"
+        fi
+
+        upgrade_script_path="$(versioned_script_path install_upgrade.escript "$newest_vsn")"
+
         ERL_FLAGS="${ERL_FLAGS:-} $EPMD_ARGS" \
-        exec "$BINDIR/escript" "$RUNNER_ROOT_DIR/bin/install_upgrade.escript" \
+        exec "$BINDIR/escript" "$upgrade_script_path" \
              "$COMMAND" "{'$REL_NAME', \"$NAME_TYPE\", '$NAME', '$COOKIE'}" "$@"
         ;;
 

bin/install_upgrade.escript

@@ -33,7 +33,7 @@ main(Args) ->
 unpack({RelName, NameTypeArg, NodeName, Cookie}, Opts) ->
     TargetNode = start_distribution(NodeName, NameTypeArg, Cookie),
     Version = proplists:get_value(version, Opts),
-    case unpack_release(RelName, TargetNode, Version) of
+    case unpack_release(RelName, TargetNode, Version, Opts) of
         {ok, Vsn} ->
             ?INFO("Unpacked successfully: ~p", [Vsn]);
         old ->
@@ -57,7 +57,7 @@ install({RelName, NameTypeArg, NodeName, Cookie}, Opts) ->
     TargetNode = start_distribution(NodeName, NameTypeArg, Cookie),
     Version = proplists:get_value(version, Opts),
     validate_target_version(Version, TargetNode),
-    case unpack_release(RelName, TargetNode, Version) of
+    case unpack_release(RelName, TargetNode, Version, Opts) of
         {ok, Vsn} ->
             ?INFO("Unpacked successfully: ~p.", [Vsn]),
             check_and_install(TargetNode, Vsn),
@@ -132,12 +132,13 @@ uninstall({_RelName, NameTypeArg, NodeName, Cookie}, Opts) ->
 uninstall(_, Args) ->
     ?INFO("uninstall: unknown args ~p", [Args]).
 
-versions({_RelName, NameTypeArg, NodeName, Cookie}, []) ->
+versions({_RelName, NameTypeArg, NodeName, Cookie}, _Opts) ->
     TargetNode = start_distribution(NodeName, NameTypeArg, Cookie),
     print_existing_versions(TargetNode).
 
 parse_arguments(Args) ->
-    parse_arguments(Args, []).
+    IsEnterprise = os:getenv("IS_ENTERPRISE") == "yes",
+    parse_arguments(Args, [{is_enterprise, IsEnterprise}]).
 
 parse_arguments([], Acc) -> Acc;
 parse_arguments(["--no-permanent"|Rest], Acc) ->
@@ -146,9 +147,10 @@ parse_arguments([VersionStr|Rest], Acc) ->
     Version = parse_version(VersionStr),
     parse_arguments(Rest, [{version, Version}] ++ Acc).
 
-unpack_release(RelName, TargetNode, Version) ->
+unpack_release(RelName, TargetNode, Version, Opts) ->
-    StartScriptExists = filelib:is_dir(filename:join(["releases", Version, "start.boot"])),
+    StartScriptExists = filelib:is_regular(filename:join(["releases", Version, "start.boot"])),
     WhichReleases = which_releases(TargetNode),
+    IsEnterprise = proplists:get_value(is_enterprise, Opts),
     case proplists:get_value(Version, WhichReleases) of
         Res when Res =:= undefined; (Res =:= unpacked andalso not StartScriptExists) ->
             %% not installed, so unpack tarball:
@@ -156,7 +158,7 @@ unpack_release(RelName, TargetNode, Version) ->
             %%      releases/<relname>-<version>.tar.gz
             %%      releases/<version>/<relname>-<version>.tar.gz
             %%      releases/<version>/<relname>.tar.gz
-            case find_and_link_release_package(Version, RelName) of
+            case find_and_link_release_package(Version, RelName, IsEnterprise) of
                 {_, undefined} ->
                     {error, release_package_not_found};
                 {ReleasePackage, ReleasePackageLink} ->
@@ -206,7 +208,7 @@ extract_tar(Cwd, Tar) ->
 %%    to the release package tarball found in 1.
 %% 3. return a tuple with the paths to the release package and
 %%    to the symlink that is to be provided to release handler
-find_and_link_release_package(Version, RelName) ->
+find_and_link_release_package(Version, RelName, IsEnterprise) ->
     RelNameStr = atom_to_list(RelName),
     %% regardless of the location of the release package, we'll
     %% always give release handler the same path which is the symlink
@@ -217,7 +219,12 @@ find_and_link_release_package(Version, RelName) ->
     %% we've found where the actual release package is located
     ReleaseLink = filename:join(["releases", Version,
                                  RelNameStr ++ ".tar.gz"]),
-    TarBalls = filename:join(["releases", RelNameStr ++ "-*" ++ Version ++ "*.tar.gz"]),
+    ReleaseNamePattern =
+        case IsEnterprise of
+            false -> RelNameStr;
+            true -> RelNameStr ++ "-enterprise"
+        end,
+    TarBalls = filename:join(["releases", ReleaseNamePattern ++ "-" ++ Version ++ "*.tar.gz"]),
     case filelib:wildcard(TarBalls) of
         [] ->
             {undefined, undefined};