DMHS主从两端数据不同步的解决办法

发表于 更新于 分类于

dmhs部署可参考官方文档：

1
https://eco.dameng.com/community/post/20221221162103LX655O1WTFKL93HLZ5

如果dmhs需要同步的数据过大且包含大字段，直接通过copy初始化可能需要数天才能完成。

可以通过dmrman物理备份数据库并同步到dmhs两个节点，然后再采集端执行：

1
COPY 0 "SCH.NAME='V3XUSER'" DICT

即可快速完成dmhs的初始化。

如果执行段启动exec报错：

1
lib libdmhs_exec.so can not found,error code 0, libdmoci.so: cannot open shared object file: No such file or directory

则需要将 libdmoci.so放到dmhs的bin目录。

1
2
cd /opt/dmhs/
cp ./bin/stat/libdmoci.so ./bin/

如果执行端遇到如下报错：

1
2
3
4
5
6
7
8
9
10
11
2023-10-19 14:41:57 EXE[ERROR]: SITEID:1 V3XUSER.JK_SCHEDULER_STATE(UPDATE): affect the number of rows (0) and the expected number of rows (1)! SEQID:10681224097 TRXID:5063332283 OP_ID:4
2023-10-19 14:41:57 EXE[ERROR]: SITEID:1 V3XUSER.JK_SCHEDULER_STATE(UPDATE) sync failed, fill rows 1, SEQID:10681224097 TRXID:5063332283 OP_ID:4
2023-10-19 14:41:57 EXE[WARN]: Transactions execution error, canceling and re execute...
2023-10-19 14:41:57 EXE[INFO]: DDL:SITEID:1 SEQID:10681226561 TRXID:445406056833092 V3XUSER.ZZZZ CREATE
2023-10-19 14:41:57 EXE[ERROR]: SITEID:1 V3XUSER.JK_SCHEDULER_STATE(UPDATE): affect the number of rows (0) and the expected number of rows (1)! SEQID:10681224087 TRXID:5063332282 OP_ID:4
2023-10-19 14:41:57 EXE[ERROR]: SITEID:1 V3XUSER.JK_SCHEDULER_STATE(UPDATE) sync failed, fill rows 1, SEQID:10681224087 TRXID:5063332282 OP_ID:4
2023-10-19 14:41:57 EXE[ERROR]: save error trx, SITEID:1 SEQID:10681224087 TRXID:5063332282 OP_ID:4(V3XUSER.JK_SCHEDULER_STATE)
2023-10-19 14:41:57 EXE[INFO]: siteid: 1, table (V3XUSER.CTP_USER_HISTORY_MESSAGE) start SEQID:0
2023-10-19 14:41:57 EXE[INFO]: siteid: 1, table (V3XUSER.ZZZZ) start SEQID:0
2023-10-19 14:41:57 EXE[INFO]: siteid: 1, table (V3XUSER.CTP_USER_MESSAGE) start SEQID:0
2023-10-19 14:41:57 PUB[ERROR]: exec transaction failed, DMHS exit

从报错可以看出是V3XUSER.JK_SCHEDULER_STATE这张表数据不同步了，可以用下面的方法解决：

在采集端执行：

1
2
3
stop cpt
COPY 0 "SCH.NAME='V3XUSER' AND TAB.NAME='JK_SCHEDULER_STATE'" TRUNCATE|INSERT|THREAD|8
start cpt

重新启动执行端exec，故障解除。

如果出现如下错误：

1
2
3
4
5
2023-10-19 14:36:15 CPT[INFO]: DM8 log format is RLOG_PKG, arch: /opt/dmdbms/arch/ARCHIVE_LOCAL1_0x320D63E6[0]_2023-10-19_14-16-46.log, db_magic: 839738342 - 839738342
2023-10-19 14:36:15 CPT[ERROR]: Database Magic is not consistent(839738342:866064215), maybe DB is restored or changed, please check connected DB or reload dictionary using CLEAR mask and delete checked file: dmhs_1_0.ini
2023-10-19 14:36:15 CPT[INFO]: DM8 CPT is free.
2023-10-19 14:36:15 MGR[ERROR]: initial log analysis module failure
2023-10-19 14:36:15 MGR[ERROR]: unable to start capture, ID no: 1

只需要删除采集端dmhs的bin目录中的dmhs_1_0.ini文件就可以了。

如果遇到错误如下：

1
2
3
4
2023-10-19 14:28:40 CPT[ERROR]: DDL_MASK is not empty, but some DDL auxiliary table is missed, please rebuild it or set DDL_MASK = 0.
2023-10-19 14:28:40 CPT[INFO]: DM8 CPT is free.
2023-10-19 14:28:40 MGR[ERROR]: initial log analysis module failure
2023-10-19 14:28:40 MGR[ERROR]: unable to start capture, ID no: 1

则是需要在采集端重新执行如下脚本：

1
/opt/dmhs/scripts/ddl_sql_dm8.sql

路径可能不通，搜索文件名即可。

如果遇到如下错误：

1
2023-10-19 14:25:51 CPT[ERROR]: DDL_MASK is not empty, but some DDL auxiliary table is missed, please rebuild it or set DDL_MASK = 0.

则是需要重新初始化：

1
COPY 0 "SCH.NAME='V3XUSER'" DICT

故障解除。