问题现象
全文索引更新,处理时报错
查看 express.log,gcluster 层有如下的异常信息:
2018-05-23 20:42:15.653 [SQLDISP][ERROR][S:3651112][Q:119671099]:Query failed, THD(0x11
506e000) HOST(15.150.8.95:5050->6138812) reason: Lost connection to GBase server during qu
ery SQL(UPDATE GCLOCAL INDEX "IX_RG010005" ON "horus"."rwa_source_0001_n6")
2018-05-23 20:42:30.716 [SQLDISP][ERROR][S:3651112][Q:119671099]:Fail to create KillConne
ction, reason: Can't connect to GBase server on '15.150.8.94(cur_time:2018-05-23 20:42:30.716
used_time:15s)' (110).
2018-05-23 20:42:45.716 [SQLDISP][ERROR][S:3651112][Q:119671099]:Fail to create KillConne
ction, reason: Can't connect to GBase server on '15.150.8.95(cur_time:2018-05-23 20:42:45.716
used_time:14s)' (110).
解决方法
全文默认参数是支持增量更新的,会在索引过程中持有字典,大批量一次更
新会导致内存吃紧,可以设置 gbfticfg.xml 配置文件中的参数 reduceMemory
Mode 为 1 解决。
全文配置如下:
cat /opt/gnode/server/lib/gbase/plugin/gbfti/cfg/gbfticfg.xml
......
<segThreads>4</segThreads>
<sortThreads>4</sortThreads>
<outThreads>3</outThreads>
<reduceMemMode>1</reduceMemMode>
GBase 8a MPP 全文索引更新处理时报错
最新推荐文章于 2024-03-20 15:32:10 发布