debug流程
java bug
项目
发布日期:   2020-12-19
作者:   锦泉
文章字数:   2.6k
阅读时长:   12 分
阅读次数:  

思路

  1. 首先就是仔细看反馈表中的问题，有些报回来的bug是没问题的，只是测试那边不了解业务流程，这种说明一下情况基本就解决了；
  2. 然后就是真的有问题的bug，先进行bug复现，先看看本地服务器会不会有这bug，如果本地服务器也报这个bug，那就是确实又问题了，这样就开始debug。如果本地服务器没问题，额这种情况就要远程连接医院那边的电脑进行bug排查。
  3. 先看报错的语句，复制到源码看看能不能找到，如果不能找到的话可能是前端的问题，报的前端的错误，也可能是本地服务器的权限中心字段名没对应上，也就是权限中心没配置好，这时就要找对应的负责人沟通了；如果能源码找到跑错的语句才开始真正的bug排查。
  4. 开始bug排查，f12打开浏览器控制台，复现bug查看控制台信息看是调用了哪个接口，再看消毒供应室的源码找到对应的rest，再找service，mapper，一步一步深入看看是那个部分哪里出错了。
  5. 如果按4这样粗略地排查看不出来的话就要看日志了，连接上本地服务器，命令行输入 tail -300f log日志文件 ，然后复现报bug的操作，拿到日志信息看具体是哪里有问题了，然后用swagger或者postman发送信息在本地环境debug一步一步排查，这种这么仔细一般都能解决大部分bug了。

举个栗子

比如我遇到了一个事务无限回滚的问题，这就的的确确是个大问题，先在本地环境尝试着能不能复现，在谷歌浏览器中按F12，查看回滚那一个操作到底干了些啥

比如上图url为：http://10.0.1.117:10079/ssd/v1.0/flow/sterilize/batch/qt

http://10.0.1.117:10079 是服务器，本地就是localhost

/ssd/v1.0 是项目路径（server.context-path）。
如果server.context-path没有配，请求的url地址就是 localhost : port/flow/function；
如果server.context-path = “/ssd/v1.0”, 请求的url地址就是 localhost : port/ssd/v1.0/flow/function

/flow/sterilize/batch/qt 才是我们在源码要找的部分

然后就去源码rest里找啦，一步一步找到rest.java文件，例如：

如果找的过程中卡住了或者找不到的话就直接快捷键 ctrl + shift + f 全局搜索。

找到@RequestMapping，发现value跟我们要找的url末尾部分完全一致，就是找对地方啦（有些目录要分两三段找，反正就是跟文件夹差不多）

然后从rest->service->mapper的顺序过一过流程看看有没有什么错误，嗯，说实话一般都看不出来，这时候就要靠强大的log日志帮我们精确定位问题所在了。

一般项目都会有日志导出的文件，用tail -100f “日志文件”找到最近100条日志信息，来看看它到底是哪里出了问题。出问题的日志如下：

Creating a new SqlSession
Registering transaction synchronization for SqlSession [org.apache.ibatis.session.defaults.DefaultSqlSession@f842db6]
JDBC Connection [com.mysql.cj.jdbc.ConnectionImpl@7b15ae54] will be managed by Spring
==>  Preparing: select a.id as "id", a.define_id as "defineId", a.status as "status", a.dept_id as "deptId", a.name as "name", a.warehouse_status as "warehouseStatus", a.invalid_time as "invalidTime", a.valid_time as "validTime", a.action as "action", c.expected_time as "expectedTime" from t_flow_packet_instacne a left join t_flow_sterilize_batch_instance b on a.id = b.packet_instance_id left join t_flow_sterilize_batch c on b.sterilize_batch_id = c.id where a.delete_flag = 0 and c.expected_time < now() and a.status = ? and a.warehouse_status = ?
==> Parameters: 1(String), 1(String)


[2020-12-31 10:54:54.499][DEBUG][http-nio-10081-exec-3][10.0.1.117][245bacd8987a4fa383a0ba6ecfb414ec][192.168.8.98][org.springframework.web.client.HttpMessageConverterExtractor..][] "Reading to [cn.wisefly.auth.rpc.bean.AuthCommonResult>]"
[2020-12-31 10:54:54.500][INFO ][http-nio-10081-exec-3][10.0.1.117][245bacd8987a4fa383a0ba6ecfb414ec][192.168.8.98][cn.wisefly.ssd.core.busi.RestAuthorizationBusi.checkAuthorization.68][checkAuthorization] "intercepter current thread:27,http-nio-10081-exec-3"
Creating a new SqlSession
SqlSession [org.apache.ibatis.session.defaults.DefaultSqlSession@43c921b8] was not registered for synchronization because synchronization is not active
[2020-12-31 10:54:54.500][DEBUG][http-nio-10081-exec-3][10.0.1.117][245bacd8987a4fa383a0ba6ecfb414ec][192.168.8.98][org.springframework.jdbc.datasource.DataSourceUtils..][] "Fetching JDBC Connection from DataSource"

SqlSession was not registered for synchronization because synchronization is not active，很显然，出问题的sql语句就是这段话前一个sql了。

竟然都找到出问题的sql语句了，就去数据库调试sql了，这个栗子为什么会无限回滚是因为一个实例对应多条数据，而读每个实例对应的数据时都只读一条，而且还是最老的数据，因为有定时过期灭菌检测，最老的数据不管怎么操作都是过期的，所以才出现无限回滚的情况。

解决方法是我们去读取最新的数据。原本sql语句为：

select a.id as "id", a.define_id as "defineId", a.status as "status", 
a.dept_id as "deptId", a.name as "name", a.warehouse_status as "warehouseStatus", 
a.invalid_time as "invalidTime", a.valid_time as "validTime", 
a.action as "action", c.expected_time as "expectedTime" 
from t_flow_packet_instacne a 
left join t_flow_sterilize_batch_instance b on a.id = b.packet_instance_id 
left join t_flow_sterilize_batch c on b.sterilize_batch_id = c.id 
where a.delete_flag = 0 
and c.expected_time < now() 
and a.status = 1 
and a.warehouse_status = 1

修改了下，加了个where判断：

select a.id as "id", a.define_id as "defineId", a.status as "status", 
a.dept_id as "deptId", a.name as "name", a.warehouse_status as "warehouseStatus", 
a.invalid_time as "invalidTime", a.valid_time as "validTime", 
a.action as "action", c.expected_time as "expectedTime" 
from t_flow_packet_instacne a 
left join t_flow_sterilize_batch_instance b on a.id = b.packet_instance_id 
left join t_flow_sterilize_batch c on b.sterilize_batch_id = c.id 
where a.delete_flag = 0 
and c.expected_time < now() 
and a.status = 1 
and a.warehouse_status = 1 
and b.last_time = (select max(b.last_time) 
from t_flow_packet_instacne a 
left join t_flow_sterilize_batch_instance b 
on a.id = b.packet_instance_id )

其实就是让它选择查找最新的数据。

就这样这个结合业务的bug就解决了！！！只是改了下sql语句而已，主要是定位问题，所以定位bug需要花些功夫。我还走了一边debug的流程，加断点，一步一步看参数，结果根本发现不了问题，因为我那时定位的地方不对，走了好几遍debug都没看出啥问题，结果原来是定时任务检测回滚了数据。

栗子后续

处理完上一个bug后，走测试时发现了一个新bug，就是更新数据同一个实例会更新出多条数据展示出来，这也是一个实例对应多条数据的问题，然后我去看数据库实例所对应的表的时候发现实例都是唯一的，说明问题还是出现在一对多的问题上。

Bug：更新实例时会刷新出该实例对应的多条数据。

目标：我们想让实例展示唯一一条数据，而且还是最新的一条数据。

跟上一个栗子操作差不多，先去F12看后台到底调用了哪个接口

然后找到对应的rest：

对应的接口代码：

rest层（FlowStorageRest）

  @GetMapping(value = "/s")
  public ResponsePacket<PageResponseData<TFlowPacketInstance>> list(HttpServletRequest request) {

    Page page = PageUtils.packagingPage(request);
    Condition condition = new Condition(request);
    if (condition.get("_sort") == null && condition.get("_order") == null) {
      condition.put("_sort", "a.create_time");
      condition.put("_order", "desc");
    }
    Set<String> warehouseStatuses = new HashSet<>();
    //hbt reset
    warehouseStatuses.add(FlowGlobalValueBean.WarehouseStatus.PUBLISH);
    warehouseStatuses.add(FlowGlobalValueBean.WarehouseStatus.NO_PUBLISH);
    warehouseStatuses.add(FlowGlobalValueBean.WarehouseStatus.SCRAP);
    warehouseStatuses.add(FlowGlobalValueBean.WarehouseStatus.OUT);
    warehouseStatuses.add(FlowGlobalValueBean.WarehouseStatus.WAIT_STERILIZE);
    condition.put("warehouseStatuses", warehouseStatuses);

    //调用service层的listFailureTime
    page = packetInstanceService.listFailureTime(condition, page);
    Map<String, Map<String, String>> dictData = dictService.loadDict("ssd_warehouse_status", "ssd_packet_status", "ssd_warehouse_action");
    PageResponseData<TFlowPacketInstance> pageResponseData = new PageResponseData<>(page, dictData);
    return ResponsePacket.generateSuccessPacket(pageResponseData);
  }

如果文件太大太长不好找接口的话，可以选择用idea快捷键ctrl + f快速找到该文件下的接口位置

service层：

  @Transactional
  public Page listFailureTime(Map<String, Object> condition, Page page) {

    if (page == null) {
      page = new Page();
    }

    int startIndex = page.getStartIndex().intValue();
    int pageSize = page.getVolumePage().intValue();

    Integer count = mapper.countFailureTime(condition);
    if (count == null || count.intValue() == 0) {
      page.setTotalCount(0);
      page.setRecords(new ArrayList());
      return page;
    }

    //调用mapper层的listFailureTime
    List<Map<String, Object>> list = mapper.listFailureTime(condition, new RowBounds(startIndex, pageSize));

    page.setTotalCount(count);
    page.setRecords(list);

    return page;
  }

mapper层：

<select id="listFailureTime" resultType="Map" parameterType="Map">
    select
    a.id as "id",
    a.define_id as "defineId",
    a.packet_data as "packetData",
    a.status as "status",
    a.dept_id as "deptId",
    a.name as "name",
    a.warehouse_status as "warehouseStatus",
    a.create_user as "createUser",
    a.create_time as "createTime",
    a.last_user as "lastUser",
    a.last_time as "lastTime",
    a.delete_flag as "deleteFlag",
    a.invalid_time as "invalidTime",
    a.valid_time as "validTime",
    a.action as "action",
    c.expected_time as "expectedTime"
    from
    <include refid="table_name"/>
    a
    left join t_flow_sterilize_batch_instance b on a.id = b.packet_instance_id
    left join t_flow_sterilize_batch c on b.sterilize_batch_id = c.id
    where a.delete_flag = 0
    group by a.id)
    <include refid="condition_sql"/>
    <include refid="failureTimeCreateUserFuzzy"></include>
    <if test=" _sort !=null and _sort != ''">
      order by ${_sort}
      <if test="_order !=null and _order != ''">
        ${_order}
      </if>
    </if>

  </select>

然后就是去本地测试数据库环境进行调试，如果不想自己拼接sql的话可以选择从导出的日志文件里边拷贝sql语句，这样就不用自己一段段拼接这么辛苦了。比如下面的sql就是我从日志中直接粘贴出来的：

select a.id as "id", a.define_id as "defineId", a.packet_data as "packetData", a.status as "status", a.dept_id as "deptId", a.name as "name", a.warehouse_status as "warehouseStatus", 
a.create_user as "createUser", a.create_time as "createTime", a.last_user as "lastUser", a.last_time as "lastTime", a.delete_flag as "deleteFlag", a.invalid_time as "invalidTime", 
a.valid_time as "validTime", a.action as "action", c.expected_time as "expectedTime" 
from t_flow_packet_instacne a 
left join t_flow_sterilize_batch_instance b on a.id = b.packet_instance_id 
left join t_flow_sterilize_batch c on b.sterilize_batch_id = c.id 
where a.delete_flag = 0 
and a.define_id= 'AF200914R3VGF3'
and a.warehouse_status in ( 1 , 2 , 3 , 4 , 6 )
order by a.create_time desc 
limit 0,20

这里因为实例对应的数据只有id是相同的，其他行大部分都不同，没办法用distinct来去重，就要用到某一字段的唯一性特点及group by分组。

# 模板
select * from table where id in (select max(id) from table group by [去除重复的字段名列表,....])

# 套模板
select max(c.expected_time) 
from t_flow_packet_instacne a 
left join t_flow_sterilize_batch_instance b on a.id = b.packet_instance_id 
left join t_flow_sterilize_batch c on b.sterilize_batch_id = c.id 
group by a.id

上面那条sql就是只保留每一个实例最新过期时间

筛选最新的数据输出（即最大的expected_time）：

select a.id as "id", a.define_id as "defineId", a.packet_data as "packetData", a.status as "status", a.dept_id as "deptId", a.name as "name", a.warehouse_status as "warehouseStatus", 
a.create_user as "createUser", a.create_time as "createTime", a.last_user as "lastUser", a.last_time as "lastTime", a.delete_flag as "deleteFlag", a.invalid_time as "invalidTime", 
a.valid_time as "validTime", a.action as "action", c.expected_time as "expectedTime" 
from t_flow_packet_instacne a 
left join t_flow_sterilize_batch_instance b on a.id = b.packet_instance_id 
left join t_flow_sterilize_batch c on b.sterilize_batch_id = c.id 
where a.delete_flag = 0 
and c.expected_time 
in (select max(c.expected_time) 
from t_flow_packet_instacne a 
left join t_flow_sterilize_batch_instance b on a.id = b.packet_instance_id 
left join t_flow_sterilize_batch c on b.sterilize_batch_id = c.id 
group by a.id)
and a.define_id= 'AF200914R3VGF3'
and a.warehouse_status in ( 1 , 2 , 3 , 4 , 6 )
order by a.create_time desc 
limit 0,20

然后把修改的部分粘贴到mapper里，install，重新发布新jar包就可以了。

到这就把坑给填了，果然处理了一个bug就会诞生更多的bug。

附送一张BUG处理流程（Boss画的）：

   转载规则


《debug流程》 由 锦泉 采用 知识共享署名 4.0 国际许可协议 进行许可。
  目录