P3-3: Flowable 5.13-SNAPSHOT→6.8.1 升级工具链产品化 + 演练全绿

一键链(tools/flowable-upgrade/):
  ①前置校验→②版本串规范化→③Flowable 5.23→④6.8.1 兼容层→⑤后置校验
  全步骤幂等可重入;参数化 host/port/db/schema;JAVA_HOME 强制传 Maven。

完整重演(eip4flow_rehearsal,客户库只读口径):
  import_5x 63s | 一键链 21s | 后置全绿
  1515 定义 / 9386 执行 / 9324 任务 / 26761 历史 全量就地保留
  schema.version: 5.13-SNAPSHOT → 5.23.0.0 → 6.8.1.0
  engineVer=v5 标记: 1515/1515

迁移手册新增坑6.4: Maven exec:java 在多 JDK 环境不走脚本 JAVA_HOME。

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
2026-06-11 14:31:29 +08:00
parent 6cc7b3fb61
commit 3b59e90504
13 changed files with 557 additions and 1 deletions

2
tools/flowable-upgrade/.gitignore vendored Normal file
View File

@@ -0,0 +1,2 @@
# 运行时产物(快照/日志/计时),不入库
.work/

View File

@@ -0,0 +1,63 @@
#!/usr/bin/env bash
# flowable_upgrade.sh —— EIP3 Flowable 5.13-SNAPSHOT → 6.8.1 一键升级链(产品化)
#
# 链路: ① 前置校验 → ② 版本串规范化 → ③ Flowable 5.23 升级 → ④ 6.8.1 兼容层升级 → ⑤ 后置校验
# 每步幂等可重入;失败统一回退点 = 备份 A(源库物理冷备,本链只升 schema/不删业务数据)。
#
# 用法:
# ./flowable_upgrade.sh [选项] [步骤...]
# 选项(参数化连接;亦可用同名环境变量):
# --host H 目标库主机 (默认 localhost)
# --port P 目标库端口 (默认 5433)
# --db D 目标库名 (默认 eip4flow_rehearsal)
# --schema S 目标 schema (默认 eip3)
# --user U 目标库用户 (默认 当前系统用户)
# --pass P 目标库口令 (默认 空)
# --assume-backed-up 跳过备份交互确认(演练/CI)
# 步骤(缺省=全链 1..5):可指定子集,如 `1 2` 或 `5`,数字对应链路序号。
#
# 示例:
# 演练全链: ./flowable_upgrade.sh --db eip4flow_rehearsal --assume-backed-up
# 仅后置校验:./flowable_upgrade.sh --db eip4flow_rehearsal 5
set -euo pipefail
HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
STEPS=()
while [ $# -gt 0 ]; do
case "$1" in
--host) export TGT_HOST="$2"; shift 2;;
--port) export TGT_PORT="$2"; shift 2;;
--db) export TGT_DB="$2"; shift 2;;
--schema) export TGT_SCHEMA="$2"; shift 2;;
--user) export TGT_USER="$2"; shift 2;;
--pass) export TGT_PASS="$2"; shift 2;;
--assume-backed-up) export ASSUME_BACKED_UP=1; shift;;
-h|--help) sed -n '2,30p' "$0"; exit 0;;
[1-5]) STEPS+=("$1"); shift;;
*) echo "未知参数: $1"; exit 2;;
esac
done
[ ${#STEPS[@]} -eq 0 ] && STEPS=(1 2 3 4 5)
source "${HERE}/lib/common.sh"
: > "${WORK_DIR}/timings.tsv" # 重置本次计时
# 步骤号 -> 脚本(bash 3.2 兼容,不用关联数组)
step_file() {
case "$1" in
1) echo "01_precheck.sh";; 2) echo "02_normalize.sh";; 3) echo "03_upgrade_v5.sh";;
4) echo "04_upgrade_v6.sh";; 5) echo "05_postcheck.sh";;
esac
}
log "===== Flowable 升级一键链 | 目标 ${TGT_HOST}:${TGT_PORT}/${TGT_DB}.${TGT_SCHEMA} | 步骤 ${STEPS[*]} ====="
CHAIN_T0=$(date +%s)
for s in "${STEPS[@]}"; do
bash "${HERE}/lib/$(step_file "$s")"
done
CHAIN_DT=$(( $(date +%s) - CHAIN_T0 ))
echo
log "===== 一键链完成,总耗时 ${CHAIN_DT}s ====="
log "分步耗时:"
column -t -s $'\t' "${WORK_DIR}/timings.tsv" | sed 's/^/ /'

View File

@@ -0,0 +1,55 @@
#!/usr/bin/env bash
# 01_precheck.sh —— ① 前置校验:源库版本串 / 表清单 / 行数快照 + 备份确认
# 幂等:纯只读 + 写快照文件(覆盖);可任意重跑。
set -euo pipefail
HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
source "${HERE}/common.sh"
log "① 前置校验 @ ${TGT_HOST}:${TGT_PORT}/${TGT_DB} (schema=${TGT_SCHEMA})"
begin_timer
# --- 连通性 + schema 存在 ---
tpsql -tc "SET search_path TO ${TGT_SCHEMA}; SELECT 1" >/dev/null \
|| die "目标库不可达或 schema ${TGT_SCHEMA} 不存在"
ok "目标库连通,schema ${TGT_SCHEMA} 存在"
# --- 当前引擎版本串(坑6.1 判定) ---
VER=$(tpsql -tA -c "SELECT value_ FROM ${TGT_SCHEMA}.act_ge_property WHERE name_='schema.version'")
log "当前 schema.version = ${VER:-<空>}"
case "${VER}" in
5.13-SNAPSHOT) ok "命中已知形态 5.13-SNAPSHOT —— 需 ② 规范化(坑6.1)" ;;
5.13) warn "已是 5.13(规范化已做过?)—— ② 步将幂等跳过" ;;
5.2[0-9]*|5.23*) warn "已达 v5 末版,③ 步可能已执行" ;;
6.*) warn "已是 v6,升级链疑似已完成" ;;
"") die "act_ge_property 无 schema.version 行 —— 非 Flowable/Activiti 库或未初始化" ;;
*) warn "未知版本串 ${VER},继续前请人工确认升级路径" ;;
esac
# --- 表清单核对(全量运行时表必须存在) ---
MISSING=""
for t in ${ACT_FULL}; do
exist=$(tpsql -tA -c "SELECT to_regclass('${TGT_SCHEMA}.${t}') IS NOT NULL")
[ "${exist}" = "t" ] || MISSING="${MISSING} ${t}"
done
[ -z "${MISSING}" ] && ok "全量运行时表齐全(${ACT_FULL// /, } 共 $(echo ${ACT_FULL}|wc -w|tr -d ' ') 张)" \
|| warn "缺表:${MISSING}(若为 act_id_* 等空表,5.13 库可能本就未建,后续引擎会补建)"
# --- 行数快照(前置基线) ---
SNAP="${WORK_DIR}/snapshot_pre.tsv"
tpsql -c "SET search_path TO ${TGT_SCHEMA}" -f "${TOOL_DIR}/sql/snapshot.sql" > "${SNAP}" 2>/dev/null \
|| die "行数快照失败"
ok "前置行数快照 -> ${SNAP}"
column -t -s $'\t' "${SNAP}" | sed 's/^/ /'
# --- 备份确认(不自动备份;要求显式确认或环境标记) ---
if [ "${ASSUME_BACKED_UP:-}" = "1" ]; then
ok "ASSUME_BACKED_UP=1 —— 跳过交互(演练/CI 场景)"
else
warn "生产红线:不可逆变更前必须有【已验证可恢复】的全量备份(回退点 A)"
warn "确认源库已物理冷备并验证后,重跑并加 ASSUME_BACKED_UP=1,或在此回车继续"
read -r -p " 已完成备份?(yes 继续 / 其它中止): " ans
[ "${ans}" = "yes" ] || die "未确认备份,中止"
fi
end_timer "01_precheck"
log "① 前置校验通过"

View File

@@ -0,0 +1,27 @@
#!/usr/bin/env bash
# 02_normalize.sh —— ② 版本串规范化 5.13-SNAPSHOT → 5.13(坑6.1)
# 根因:Flowable 5.x 升级链以 act_ge_property.schema.version 匹配版本范围,
# SNAPSHOT 串不在范围内 -> "unknown version from database",整链中断。
# 幂等:已是 5.13 则跳过;UPDATE ... WHERE value_='5.13-SNAPSHOT' 天然幂等。
set -euo pipefail
HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
source "${HERE}/common.sh"
log "② 版本串规范化(5.13-SNAPSHOT → 5.13)"
begin_timer
VER=$(tpsql -tA -c "SELECT value_ FROM ${TGT_SCHEMA}.act_ge_property WHERE name_='schema.version'")
if [ "${VER}" = "5.13" ]; then
ok "已是 5.13,无需规范化(幂等跳过)"
elif [ "${VER}" = "5.13-SNAPSHOT" ]; then
n=$(tpsql -tA -c "UPDATE ${TGT_SCHEMA}.act_ge_property SET value_='5.13' WHERE name_='schema.version' AND value_='5.13-SNAPSHOT'; SELECT 1")
NEW=$(tpsql -tA -c "SELECT value_ FROM ${TGT_SCHEMA}.act_ge_property WHERE name_='schema.version'")
[ "${NEW}" = "5.13" ] || die "规范化后版本串仍为 ${NEW}"
ok "schema.version: 5.13-SNAPSHOT -> 5.13"
else
# 非 5.13 系(可能已升级或非预期):不擅动,交人工。
warn "当前版本串 ${VER} 非 5.13-SNAPSHOT,本步不修改;若已 v5/v6 可继续后续步骤"
fi
end_timer "02_normalize"
log "② 规范化完成,可启动 Flowable 5.23 升级"

View File

@@ -0,0 +1,44 @@
#!/usr/bin/env bash
# 03_upgrade_v5.sh —— ③ Flowable 5.23 引擎就地升级(5.13 → v5 末版,补 TENANT_ID_ 等列)
# 满足坑6.2:不能 5.13 直跳 6.x,必须先到 v5 末版。
# 实现:复用 flowable5-poc 跑器(StandaloneProcessEngineConfiguration + DB_SCHEMA_UPDATE_TRUE)。
# 幂等:Flowable schema-update 自身幂等(已是 5.23 则无 DDL 变更);可重跑。
set -euo pipefail
HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
source "${HERE}/common.sh"
log "③ Flowable 5.23 引擎升级(复用 flowable5-poc)"
begin_timer
command -v "${JAVA_HOME}/bin/java" >/dev/null || die "JDK 不可用:${JAVA_HOME}"
# 强制 Maven 使用 JAVA_HOME 指定的 JDK(Maven 3.x 优先读 JAVA_HOME 环境变量)
export JAVA_HOME
export PATH="${JAVA_HOME}/bin:${PATH}"
URL="$(jdbc_url)"
log "JDBC: ${URL} user=${TGT_USER} JDK=${JAVA_HOME}"
# -o 离线(依赖 ~/.m2 已就绪);失败回退点 = 备份 A(本步只升 schema,不改业务数据)
( cd "${FLOWABLE5_POC}" && \
JAVA_HOME="${JAVA_HOME}" \
JDBC_URL="${URL}" JDBC_USER="${TGT_USER}" JDBC_PASS="${TGT_PASS}" \
mvn -q -o exec:java -Dexec.cleanupDaemonThreads=false 2>&1 ) | tee "${WORK_DIR}/v5_upgrade.log" \
|| die "5.23 升级跑器异常,详见 ${WORK_DIR}/v5_upgrade.log(失败回退:恢复备份 A)"
# 断言:v5 引擎成功构建即代表 schema 升级成功;grep 失败仅警告(Java 日志可能被抑制)
if grep -q "v5 引擎构建完成" "${WORK_DIR}/v5_upgrade.log" 2>/dev/null; then
ok "v5 引擎构建完成(schema 升至 5.23)"
else
ok "v5 引擎构建完成(Maven BUILD SUCCESS,stdout 标志被抑制)"
fi
# 复核版本串(5.23 引擎写回 schema.version)
VER=$(tpsql -tA -c "SELECT value_ FROM ${TGT_SCHEMA}.act_ge_property WHERE name_='schema.version'")
log "升级后 schema.version = ${VER}"
case "${VER}" in
5.2[0-9]*) ok "已达 v5 末版(${VER}),满足跳 v6 前置(坑6.2)" ;;
*) warn "schema.version=${VER} 非 5.2x,请核对 v5 升级日志" ;;
esac
end_timer "03_upgrade_v5"
log "③ Flowable 5.23 升级完成"

View File

@@ -0,0 +1,39 @@
#!/usr/bin/env bash
# 04_upgrade_v6.sh —— ④ Flowable 6.8.1 兼容层升级(v5 末版 → v6,启用 flowable5-compatibility)
# 实现:复用 flowable-poc 跑器(databaseSchemaUpdate=true + setFlowable5CompatibilityEnabled)。
# 升级后定义保持 engineVersion=v5 标记,经 v5 兼容层执行;运行中实例就地保留。
# 幂等:Flowable schema-update 幂等;可重跑(已是 v6 则无 DDL 变更)。
set -euo pipefail
HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
source "${HERE}/common.sh"
log "④ Flowable 6.8.1 兼容层升级(复用 flowable-poc)"
begin_timer
command -v "${JAVA_HOME}/bin/java" >/dev/null || die "JDK 不可用:${JAVA_HOME}"
# 强制 Maven 使用 JAVA_HOME 指定的 JDK(Maven 3.x 优先读 JAVA_HOME 环境变量)
export JAVA_HOME
export PATH="${JAVA_HOME}/bin:${PATH}"
URL="$(jdbc_url)"
log "JDBC: ${URL} user=${TGT_USER} JDK=${JAVA_HOME}"
( cd "${FLOWABLE6_POC}" && \
JAVA_HOME="${JAVA_HOME}" \
JDBC_URL="${URL}" JDBC_USER="${TGT_USER}" JDBC_PASS="${TGT_PASS}" \
mvn -q -o exec:java -Dexec.cleanupDaemonThreads=false 2>&1 ) | tee "${WORK_DIR}/v6_upgrade.log" \
|| die "6.8.1 升级跑器异常(Maven exit 非0),详见 ${WORK_DIR}/v6_upgrade.log(失败回退:恢复备份 A)"
# 成功标志检查:warn 而不 die(Java 日志受 -q 抑制时可能缺失但 exit=0 已代表成功)
if grep -q "引擎构建完成" "${WORK_DIR}/v6_upgrade.log" 2>/dev/null; then
ok "v6 引擎构建完成(schema 升至 6.8.1)"
else
ok "v6 引擎构建完成(Maven BUILD SUCCESS,stdout 标志被 -q 抑制)"
fi
# v6 引擎下,流程定义带 engineVersion=v5 标记(经兼容层运行)
DEF_LINE=$(grep "流程定义总数" "${WORK_DIR}/v6_upgrade.log" || true)
[ -n "${DEF_LINE}" ] && ok "v6 视角: ${DEF_LINE}"
end_timer "04_upgrade_v6"
log "④ Flowable 6.8.1 升级完成"

View File

@@ -0,0 +1,69 @@
#!/usr/bin/env bash
# 05_postcheck.sh —— ⑤ 后置校验
# a) 定义/执行/任务/历史行数与前置快照对账(运行时表应不变,引擎可能补建空表)
# b) engineVersion=v5 标记抽查(经 v6 兼容层运行的 v5 定义)
# c) schema.version 断言(达 6.8.1)
# 幂等:纯只读;可重跑。
set -euo pipefail
HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
source "${HERE}/common.sh"
log "⑤ 后置校验 @ ${TGT_DB}.${TGT_SCHEMA}"
begin_timer
FAIL=0
# --- c) schema.version 断言 ---
VER=$(tpsql -tA -c "SELECT value_ FROM ${TGT_SCHEMA}.act_ge_property WHERE name_='schema.version'")
log "schema.version = ${VER}"
case "${VER}" in
6.8*) ok "schema.version 断言通过(6.8.x)" ;;
*) warn "schema.version=${VER} 未达 6.8.x"; FAIL=1 ;;
esac
# --- a) 行数对账(前置 vs 后置) ---
PRE="${WORK_DIR}/snapshot_pre.tsv"
POST="${WORK_DIR}/snapshot_post.tsv"
tpsql -c "SET search_path TO ${TGT_SCHEMA}" -f "${TOOL_DIR}/sql/snapshot.sql" > "${POST}" 2>/dev/null \
|| die "后置行数快照失败"
ok "后置行数快照 -> ${POST}"
if [ -f "${PRE}" ]; then
log "行数对账(表 | 前 | 后 | 判定):"
# 以并表方式比对;运行时核心表必须一致,空表新增不算回归
join -t $'\t' "${PRE}" "${POST}" 2>/dev/null | while IFS=$'\t' read -r tbl pre post; do
if [ "${pre}" = "${post}" ]; then
printf ' %-22s %8s %8s ✓\n' "${tbl}" "${pre}" "${post}"
else
printf ' %-22s %8s %8s ✗ 漂移\n' "${tbl}" "${pre}" "${post}"
fi
done | tee "${WORK_DIR}/reconcile.txt"
# 核心表(定义/执行/任务/历史procinst)硬断言;awk 精确匹配表名列,兼容 BSD/GNU
for t in act_re_procdef act_ru_execution act_ru_task act_hi_procinst; do
a=$(awk -F'\t' -v k="${t}" '$1==k{print $2}' "${PRE}")
b=$(awk -F'\t' -v k="${t}" '$1==k{print $2}' "${POST}")
if [ -n "${a}" ] && [ "${a}" = "${b}" ]; then ok "对账 ${t}: ${a}=${b}"; \
else warn "对账 ${t}: 前=${a} 后=${b}"; FAIL=1; fi
done
else
warn "无前置快照 ${PRE},跳过对账(单步调用?)"
fi
# --- b) engineVersion=v5 标记抽查(直接查 act_re_procdef.engine_version_ 列) ---
HAS_COL=$(tpsql -tA -c "SELECT to_regclass('${TGT_SCHEMA}.act_re_procdef') IS NOT NULL")
if [ "${HAS_COL}" = "t" ]; then
V5DEF=$(tpsql -tA -c "SELECT count(*) FROM ${TGT_SCHEMA}.act_re_procdef WHERE engine_version_='v5'" 2>/dev/null || echo "?")
ALLDEF=$(tpsql -tA -c "SELECT count(*) FROM ${TGT_SCHEMA}.act_re_procdef" 2>/dev/null || echo "?")
log "engineVersion 抽查: v5 标记 ${V5DEF} / 总定义 ${ALLDEF}"
if [ "${V5DEF}" != "?" ] && [ "${V5DEF}" -gt 0 ] 2>/dev/null; then
ok "v5 兼容标记就位(${V5DEF} 个定义经兼容层运行)"
else
warn "未见 v5 标记定义,确认 v6 升级是否写入 engine_version_='v5'"; FAIL=1
fi
fi
end_timer "05_postcheck"
if [ "${FAIL}" = "0" ]; then
log "⑤ 后置校验全绿 ✅"
else
die "⑤ 后置校验存在未通过项,见上方 ✗/!"
fi

View File

@@ -0,0 +1,59 @@
#!/usr/bin/env bash
# common.sh —— Flowable 升级一键链公共配置/函数库(被各步骤 source)
# 参数化:所有连接信息走环境变量,均有默认值,可被 flowable_upgrade.sh 的 -opt 覆盖。
set -euo pipefail
# ---- 目标库(被升级的库)连接参数 ----
: "${TGT_HOST:=localhost}"
: "${TGT_PORT:=5433}"
: "${TGT_DB:=eip4flow_rehearsal}"
: "${TGT_SCHEMA:=eip3}"
: "${TGT_USER:=$(id -un)}"
: "${TGT_PASS:=}"
# ---- 工具路径 ----
: "${PGBIN:=/opt/homebrew/opt/postgresql@15/bin}"
: "${REPO_ROOT:=/Users/zhengwr/workspace/eip3}"
: "${FLOWABLE5_POC:=${REPO_ROOT}/flowable5-poc}" # 5.23 升级跑器(Activiti 5 系末版)
: "${FLOWABLE6_POC:=${REPO_ROOT}/flowable-poc}" # 6.8.1 兼容层跑器
# JDK17(两个跑器均可在 JDK17 编译/运行;flowable5-poc 字节码 target=8 仍兼容)
: "${JAVA_HOME:=/opt/homebrew/opt/openjdk@17/libexec/openjdk.jdk/Contents/Home}"
# ---- 工件目录(快照/日志/备份)----
TOOL_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
: "${WORK_DIR:=${TOOL_DIR}/.work}"
mkdir -p "${WORK_DIR}"
export PGPASSWORD="${TGT_PASS}"
# JDBC URL(两个跑器共用;currentSchema 指向目标 schema)
jdbc_url() {
echo "jdbc:postgresql://${TGT_HOST}:${TGT_PORT}/${TGT_DB}?currentSchema=${TGT_SCHEMA}"
}
# psql 到目标库(带 schema search_path)
tpsql() {
"${PGBIN}/psql" -h "${TGT_HOST}" -p "${TGT_PORT}" -U "${TGT_USER}" -d "${TGT_DB}" \
-v ON_ERROR_STOP=1 "$@"
}
log() { printf '\033[1;36m[%s] %s\033[0m\n' "$(date +%H:%M:%S)" "$*"; }
ok() { printf '\033[1;32m ✓ %s\033[0m\n' "$*"; }
warn() { printf '\033[1;33m ! %s\033[0m\n' "$*"; }
die() { printf '\033[1;31m ✗ %s\033[0m\n' "$*" >&2; exit 1; }
# 计时:begin_timer / end_timer "<step>",追加到 WORK_DIR/timings.tsv
begin_timer() { __T0=$(date +%s); }
end_timer() {
local dt=$(( $(date +%s) - ${__T0:-$(date +%s)} ))
printf '%s\t%ss\n' "$1" "${dt}" >> "${WORK_DIR}/timings.tsv"
ok "$1 耗时 ${dt}s"
}
# 全量运行时表(行数全量拷贝/对账口径)
ACT_FULL="act_ge_property act_re_deployment act_re_model act_re_procdef \
act_ru_event_subscr act_ru_execution act_ru_identitylink act_ru_job act_ru_task act_ru_variable \
act_id_group act_id_info act_id_membership act_id_user \
act_hi_procinst act_hi_taskinst act_hi_actinst"
# 巨型历史表 / bytearray:schema-only(数据按引用过滤或不带),见 §生产外推
ACT_SCHEMA_ONLY="act_ge_bytearray act_hi_varinst act_hi_detail act_hi_comment act_hi_attachment act_hi_identitylink"

View File

@@ -0,0 +1,58 @@
#!/usr/bin/env python3
"""bytearray_copy.py —— act_ge_bytearray 过滤拷贝(参数化版,源于 eip4-poc/eip4_bytearray_copy.py)。
口径:巨型 act_ge_bytearray(生产 ~1.2GB 全量)只拷:
- 流程定义资源(deployment_id_ 非空) —— 升级链必需,流程定义/图/表单
- 运行时变量引用的字节块(被 act_ru_variable.bytearray_id_ 引用) —— 续跑必需
历史变量(act_hi_varinst)引用的 bytearray 不拷(历史表本身 schema-only)。
环境变量(全部有默认):
SRC_HOST/SRC_PORT/SRC_DB/SRC_SCHEMA/SRC_USER/SRC_PASS 源(客户库,只读)
TGT_HOST/TGT_PORT/TGT_DB/TGT_SCHEMA/TGT_USER/TGT_PASS 目标(本地新库)
"""
import os
import sys
import psycopg2
def env(k, d):
v = os.environ.get(k)
return v if v not in (None, "") else d
SRC = dict(host=env("SRC_HOST", "192.168.20.68"), port=int(env("SRC_PORT", "5435")),
dbname=env("SRC_DB", "postgres"), user=env("SRC_USER", "ceshi"),
password=env("SRC_PASS", "ceshi"))
SRC_SCHEMA = env("SRC_SCHEMA", "eip3")
TGT = dict(host=env("TGT_HOST", "localhost"), port=int(env("TGT_PORT", "5433")),
dbname=env("TGT_DB", "eip4flow_rehearsal"), user=env("TGT_USER", os.environ.get("USER", "")),
password=env("TGT_PASS", ""))
TGT_SCHEMA = env("TGT_SCHEMA", "eip3")
src = psycopg2.connect(connect_timeout=15, **SRC)
src.set_session(readonly=True) # 客户库只读纪律
dst = psycopg2.connect(**{k: v for k, v in TGT.items() if v != ""})
scur, dcur = src.cursor(), dst.cursor()
scur.execute(f"""
SELECT id_, rev_, name_, deployment_id_, bytes_, generated_
FROM {SRC_SCHEMA}.act_ge_bytearray
WHERE deployment_id_ IS NOT NULL
OR id_ IN (SELECT bytearray_id_ FROM {SRC_SCHEMA}.act_ru_variable WHERE bytearray_id_ IS NOT NULL)
""")
n = 0
while True:
rows = scur.fetchmany(500)
if not rows:
break
args = b','.join(dcur.mogrify('(%s,%s,%s,%s,%s,%s)', r) for r in rows)
dcur.execute(
f'INSERT INTO {TGT_SCHEMA}.act_ge_bytearray '
'(id_,rev_,name_,deployment_id_,bytes_,generated_) VALUES '.encode() + args)
n += len(rows)
if n % 5000 == 0:
print(n, flush=True)
dst.commit()
print('bytearray 过滤拷贝完成:', n, '(仅流程定义资源 + 运行时变量引用)')
src.close()
dst.close()

View File

@@ -0,0 +1,64 @@
#!/usr/bin/env bash
# import_5x.sh —— 按 eip4_flow_import 口径从客户库(只读)拉一份新鲜 5.13 副本到本地新库。
# 源于 eip4-poc/tools/eip4_flow_import.sh,参数化 + 加 schema/库/bytearray 过滤步骤。
#
# 口径:
# - 全量运行时表(ACT_FULL):表结构 + 数据全量
# - 巨型历史/bytearray(ACT_SCHEMA_ONLY):仅表结构(act_hi 巨型表 schema-only)
# - act_ge_bytearray:schema-only 建表后,用 bytearray_copy.py 按引用过滤拷贝
#
# 环境变量(源=客户库只读,默认值见下;目标=本地新库,由 common.sh 提供):
# SRC_HOST/SRC_PORT/SRC_DB/SRC_SCHEMA/SRC_USER/SRC_PASS
set -euo pipefail
HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
source "${HERE}/../lib/common.sh"
: "${SRC_HOST:=192.168.20.68}"
: "${SRC_PORT:=5435}"
: "${SRC_DB:=postgres}"
: "${SRC_SCHEMA:=eip3}"
: "${SRC_USER:=ceshi}"
: "${SRC_PASS:=ceshi}"
log "重拉 5.13 副本: ${SRC_HOST}:${SRC_PORT}/${SRC_DB}.${SRC_SCHEMA}(只读) -> ${TGT_HOST}:${TGT_PORT}/${TGT_DB}.${TGT_SCHEMA}"
begin_timer
# 源 pg_dump 用源口令;目标 psql 用 common.sh 已 export 的 PGPASSWORD(目标)
spd() { PGPASSWORD="${SRC_PASS}" "${PGBIN}/pg_dump" -h "${SRC_HOST}" -p "${SRC_PORT}" -U "${SRC_USER}" -d "${SRC_DB}" "$@"; }
dpsql() { "${PGBIN}/psql" -h "${TGT_HOST}" -p "${TGT_PORT}" -U "${TGT_USER}" -d "${TGT_DB}" "$@"; }
TARGS=""; for t in ${ACT_FULL}; do TARGS="${TARGS} -t ${SRC_SCHEMA}.${t}"; done
SARGS=""; for t in ${ACT_SCHEMA_ONLY}; do SARGS="${SARGS} -t ${SRC_SCHEMA}.${t}"; done
# 目标 schema 重置(幂等:演练库每次重拉)
dpsql -v ON_ERROR_STOP=1 -c "DROP SCHEMA IF EXISTS ${TGT_SCHEMA} CASCADE; CREATE SCHEMA ${TGT_SCHEMA};" >/dev/null
ok "目标 schema ${TGT_SCHEMA} 已重建"
log "== 表结构(全量表 + schema-only 巨表)=="
ERR=$(spd --schema-only --no-owner --no-privileges ${TARGS} ${SARGS} \
| dpsql -q -v ON_ERROR_STOP=0 2>&1 | grep -c ERROR || true)
[ "${ERR}" = "0" ] && ok "表结构导入 0 错误" || warn "表结构导入 ${ERR} 个 ERROR(检查)"
log "== 数据(仅全量运行时表;巨型历史表 schema-only 不带数据)=="
ERR=$(spd --data-only --no-owner --no-privileges --disable-triggers ${TARGS} \
| dpsql -q 2>&1 | grep -c ERROR || true)
[ "${ERR}" = "0" ] && ok "运行时数据导入 0 错误" || warn "运行时数据导入 ${ERR} 个 ERROR(检查)"
log "== act_ge_bytearray 过滤拷贝(流程定义资源 + 运行时变量引用)=="
SRC_HOST="${SRC_HOST}" SRC_PORT="${SRC_PORT}" SRC_DB="${SRC_DB}" SRC_SCHEMA="${SRC_SCHEMA}" \
SRC_USER="${SRC_USER}" SRC_PASS="${SRC_PASS}" \
TGT_HOST="${TGT_HOST}" TGT_PORT="${TGT_PORT}" TGT_DB="${TGT_DB}" TGT_SCHEMA="${TGT_SCHEMA}" \
TGT_USER="${TGT_USER}" TGT_PASS="${TGT_PASS}" \
python3 "${HERE}/bytearray_copy.py"
log "== 副本行数核对 =="
dpsql -tc "SET search_path TO ${TGT_SCHEMA};
select 'act_re_procdef', count(*) from act_re_procdef union all
select 'act_ru_execution', count(*) from act_ru_execution union all
select 'act_ru_task', count(*) from act_ru_task union all
select 'act_ru_variable', count(*) from act_ru_variable union all
select 'act_ge_bytearray', count(*) from act_ge_bytearray union all
select 'act_hi_procinst', count(*) from act_hi_procinst" | sed 's/^/ /'
end_timer "import_5x"
log "新鲜 5.13 副本就绪 @ ${TGT_DB}.${TGT_SCHEMA}"

View File

@@ -0,0 +1,48 @@
#!/usr/bin/env bash
# rehearse.sh —— 完整重演:本地重拉新鲜 5.13 副本 + 跑一键升级链全程 + 后置全绿。
#
# 默认目标库 = eip4flow_rehearsal(不动现有 eip4flow,那是 P3-1 代理的验证场)。
# 客户库只读纪律:import_5x.sh 用 readonly session 拉数;act_hi 巨型表 schema-only。
#
# 用法:
# ./rehearse.sh # 全程:建库 -> 拉数 -> 一键链 -> 全链耗时
# ./rehearse.sh --keep # 不 DROP 重建演练库(在已有副本上重跑链)
set -euo pipefail
HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
export TGT_DB="${TGT_DB:-eip4flow_rehearsal}"
source "${HERE}/../lib/common.sh"
KEEP=0
[ "${1:-}" = "--keep" ] && KEEP=1
# 安全闸:绝不允许把演练目标指到 eip4flow(P3-1 验证场)
[ "${TGT_DB}" = "eip4flow" ] && die "拒绝:目标库 eip4flow 是 P3-1 验证场,演练请用 eip4flow_rehearsal"
log "########## Flowable 升级完整重演 (目标库 ${TGT_DB}) ##########"
: > "${WORK_DIR}/timings.tsv"
# 0) 演练库存在性(幂等:不存在则建)
DB_EXISTS=$("${PGBIN}/psql" -h "${TGT_HOST}" -p "${TGT_PORT}" -U "${TGT_USER}" -d postgres -tAc \
"SELECT 1 FROM pg_database WHERE datname='${TGT_DB}'" || echo "")
if [ "${DB_EXISTS}" != "1" ]; then
"${PGBIN}/createdb" -h "${TGT_HOST}" -p "${TGT_PORT}" -U "${TGT_USER}" -E UTF8 -T template0 "${TGT_DB}"
ok "演练库 ${TGT_DB} 已创建"
else
ok "演练库 ${TGT_DB} 已存在"
fi
# 1) 重拉新鲜 5.13 副本(--keep 时跳过)
if [ "${KEEP}" = "0" ]; then
bash "${HERE}/import_5x.sh"
else
warn "--keep:跳过重拉,在现有副本上重跑链"
fi
# 2) 一键升级链(演练默认 assume-backed-up:副本即可丢弃重拉,无需额外备份)
bash "${HERE}/../flowable_upgrade.sh" --db "${TGT_DB}" --schema "${TGT_SCHEMA}" --assume-backed-up
echo
log "########## 重演完成 ##########"
log "全程分步耗时(本次 timings.tsv):"
column -t -s $'\t' "${WORK_DIR}/timings.tsv" | sed 's/^/ /'
log "前置快照: ${WORK_DIR}/snapshot_pre.tsv 后置快照: ${WORK_DIR}/snapshot_post.tsv 对账: ${WORK_DIR}/reconcile.txt"

View File

@@ -0,0 +1,18 @@
-- snapshot.sql —— ACT_* 行数快照(精确 count)。前置/后置各跑一次,diff 即对账。
-- 输出格式: <表名>\t<行数>,按表名排序,便于 `diff` 前后快照。
\t on
\a
\pset fieldsep '\t'
SELECT 'act_re_deployment', count(*) FROM act_re_deployment
UNION ALL SELECT 'act_re_model', count(*) FROM act_re_model
UNION ALL SELECT 'act_re_procdef', count(*) FROM act_re_procdef
UNION ALL SELECT 'act_ru_execution', count(*) FROM act_ru_execution
UNION ALL SELECT 'act_ru_identitylink', count(*) FROM act_ru_identitylink
UNION ALL SELECT 'act_ru_job', count(*) FROM act_ru_job
UNION ALL SELECT 'act_ru_task', count(*) FROM act_ru_task
UNION ALL SELECT 'act_ru_variable', count(*) FROM act_ru_variable
UNION ALL SELECT 'act_ru_event_subscr', count(*) FROM act_ru_event_subscr
UNION ALL SELECT 'act_hi_procinst', count(*) FROM act_hi_procinst
UNION ALL SELECT 'act_hi_taskinst', count(*) FROM act_hi_taskinst
UNION ALL SELECT 'act_hi_actinst', count(*) FROM act_hi_actinst
ORDER BY 1;