hive 慢sql 查询

hive 慢sql 查询

1. 查找 hive 执行日志存储路径（一般是 hive-audit.log ）

   比如：/var/log/Bigdata/audit/hive/hiveserver/hive-audit.log

2. 解析日志 获取 执行时间 执行 OperationId 执行人 UserName=root 执行sql 数据分隔符为 \001 并写入 hivesql.txt

   tail  -1000f  /var/log/Bigdata/audit/hive/hiveserver/hive-audit.log |grep  OperationId= |awk -F"|"  '{print $1,$4}'  |awk -F "\t" -v OFS="\001" '{print $1,$2,$6}' >> /home/yx_test/hiveSql/hivesql.txt
   1

   解析结果如下
   2024-04-03 08:37:57,851   OperationId=e0c496d6-6979-4e44-a9a3-1ec3ac2a6767UserName=rootstmt={sql语句}
   1
   2

3. 上传相关解析日志到hive

   hadoop  fs -put /home/yx_test/hiveSql/hivesql*.txt  hive/warehouse/yx_test/ods_format_datas1/tb=hivesql/
   1.刷新hive元数据
   MSCK REPAIR TABLE yx_test.ods_format_datas1
   2.然后再次解析日志 
    获取： 执行id  执行结束时间  执行开始时间 执行时长 执行人 写入表 执行sql语句
    SELECT k2.id,max_ds,min_ds,df_ds,username,insert_tb,sql from  (
   SELECT id,max_ds,min_ds,df_ds  from  (
   SELECT 
    id -- 执行id
    ,max(substr(ds,1,19)) max_ds -- 执行结束时间
    ,min(substr(ds,1,19)) min_ds -- 执行开始时间
   , unix_timestamp(max(substr(ds,1,19)))-unix_timestamp(min(substr(ds,1,19))) df_ds -- 执行时长
   from  (
   SELECT split(c1,'   OperationId=')[0] ds ,split(c1,'   OperationId=')[1] id  from  yx_test. ods_format_datas1
   where tb='hivesql' 
   -- and  c2 like '%insert%'
   )k group by id
   )k1 
   where df_ds>=600 -- 获取执行时长超过 10分钟的sql信息 
   order    by  df_ds  
   )k2 
   inner join
   -- 通过执行 id 匹配出 执行的详细sql 以及写入 表 
   (SELECT 
   id,username
   ,trim(substr(substr(lower(c3),beg,ends-beg),1, if(instr(substr(lower(c3),beg,ends-beg),' partition')!=0,instr(substr(lower(c3),beg,ends-beg),' partition'),1000))) insert_tb -- 写入表
   ,lower(c3) sql from  (
   SELECT  
   split(c1,'   OperationId=')[1] id -- 执行id
   ,c2 username -- 执行角色
   ,c3 -- 执行的sql语句
   --提取写入表前后位置
   ,if(instr(lower(c3),'table')=0,instr(lower(c3),'into')+length('into'),instr(lower(c3),'table')+length('table')) beg 
   ,instr(lower(c3),' select') ends  
   from  yx_test.ods_format_datas1
   where tb='hivesql' and  c3 like '%insert%' -- 只获取 包含  insert 的日志
   )k 
   )k3 on k2.id=k3.id
   ORDER BY   df_ds desc 
   ;
   3. 通过写入表 查看 涉及到的具体库
   SELECT * from  (
   SELECT k.DB_ID,k.`NAME`,k1.TBL_NAME from  (
   SELECT DB_ID,`NAME` FROM `dbs`
   )k 
   left join 
   (SELECT DB_ID,TBL_NAME from  tbls
   )k1 on k.DB_ID=k1.DB_ID
   )k2 WHERE TBL_NAME in ('dws_koi_role_details_day')
   
   1
   2
   3
   4
   5
   6
   7
   8
   9
   10
   11
   12
   13
   14
   15
   16
   17
   18
   19
   20
   21
   22
   23
   24
   25
   26
   27
   28
   29
   30
   31
   32
   33
   34
   35
   36
   37
   38
   39
   40
   41
   42
   43
   44
   45
   46
   47
   48
   49
   50

   效果如图：

   

部署脚本

-- 日志采集
ssh omm@192.168.0.183 'bash -s' << 'EOF'
source /opt/Bigdata/client/bigdata_env
touch /home/yangxiong/hiveSql/hivesql2.txt
hadoop  fs -put /home/yangxiong/hiveSql/hivesql*.txt obs://youkia-koi/hive/warehouse/yx_test/ods_format_datas1/tb=hivesql/
echo  >  /home/yangxiong/hiveSql/hivesql*.txt
ps -ef |grep 'tail -1000f /var/log/Bigdata/audit/hive/hiveserver/hive-audit.log' |awk '{print $2}'|xargs kill -9
nohup tail   -1000f  /var/log/Bigdata/audit/hive/hiveserver/hive-audit.log |grep  OperationId= |awk -F"|"  '{print $1,$4}'  |awk -F "\t" -v OFS="\001" '{print $1,$2,$6}' >> /home/yangxiong/hiveSql/hivesql2.txt 2>&1 &
exit
# 这里可以添加更多的命令
EOF


-- 日志解析
insert overwrite table sgz_game_common.hive_timeout_sql 
SELECT k2.id sql_id,min_ds begint_ds,max_ds end_ds,df_ds sustain_ds ,username,insert_tb,sql,'${hiveconf:ds}' ds from  (
SELECT id,max_ds,min_ds,df_ds  from  (
SELECT 
 id
 ,max(substr(ds,1,19)) max_ds
 ,min(substr(ds,1,19)) min_ds
, unix_timestamp(max(substr(ds,1,19)))-unix_timestamp(min(substr(ds,1,19))) df_ds
from  (
SELECT split(c1,'   OperationId=')[0] ds ,split(c1,'   OperationId=')[1] id  from  yx_test.ods_format_datas1
where tb='hivesql' 
-- and  c2 like '%insert%'
)k group by id
)k1 
where df_ds>=1800
order    by  df_ds  
)k2 
inner join
(
SELECT 
id,username
,trim(substr(substr(lower(c3),beg,ends-beg),1, if(instr(substr(lower(c3),beg,ends-beg),' partition')!=0,instr(substr(lower(c3),beg,ends-beg),' partition'),1000))) insert_tb
,lower(c3) sql from  (
SELECT  split(c1,'   OperationId=')[1] id,c2 username,c3
,if(instr(lower(c3),'table')=0,instr(lower(c3),'into')+length('into'),instr(lower(c3),'table')+length('table')) beg
,instr(lower(c3),' select') ends
from  yx_test.ods_format_datas1
where tb='hivesql' and  c3 like '%insert%'
)k 
)k3 on k2.id=k3.id 
-- ORDER BY   df_ds desc 
UNION all
SELECT sql_id,begint_ds,end_ds,sustain_ds,username,insert_tb,sql,ds from  (
SELECT sql_id,begint_ds,end_ds,sustain_ds,username,insert_tb,sql,ds
,row_number() OVER (PARTITION by 1=1   ORDER BY ds desc) rk
from  sgz_game_common.hive_timeout_sql
WHERE ds!='{hiveconf:ds}'
)k where rk<=30

;
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54