[IOTDB-2723] [compaction + delete] Merge recovery lost data

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.13.0-SNAPSHOT
• Fix Version/s: 0.13.0
• Component/s: Core/Compaction
• Labels: pull-request-available
• Sprint: 2022-3-Dragon
• Assignee: Liuxuxin
• Reporter: 刘珍
• Due: 18/Mar/22
• Created: 11/Mar/22 10:15
• Updated: 11/May/22 02:14
• Resolved: 15/Mar/22 13:45

Description

iotdb版本0310 f133c92e202508e9bec4294cf56f34fc4e67ff9a
问题描述：
顺序空间内合并的时候，100个设备顺序执行delete 2条记录，执行完后，
killall -u 当前用户，备份data，继续在这个data上重启iotdb，执行100个设备的count查询，
查询结果有20个设备存在少数据的现象，每个序列的期望结果是199999998点。

操作流程：
1. 0.12.5-pre4 关合并，生成数据备份，记作备份1。
在192.168.10.68的/data/liuzhen_test/1205_pre4/data_bk_close-comp
2.copy 备份1 到68机器的/data/liuzhen_test/0310_master_f133c92
2022-03-11 13:40:13,204启动成功
2022-03-11 14:23:45,657 启动delete脚本（此时有顺序空间内合并）
cat delete_when_comp.sh
#!/bin/bash
base_db_dir="/data/liuzhen_test/0309_master_iotdb-2693"
out_dev="dev_name.txt"
function expDevNameToFile()
{
${base_db_dir}/sbin/start-cli.sh -e "set max_display_num=0 ;show devices;"|grep root|awk -F '|' '

{print $2}

'>${out_dev}
}

expDevNameToFile

1. delete 2 row 1 of seq ,1 of unseq
   function del2rows()
   {
   while read LINE
   do
   ${base_db_dir}/sbin/start-cli.sh -e " select s_0 from "$LINE" where time=2024-05-24T18:00:05.000+08:00 or time=2034-01-24T18:53:22.000+08:00" >> sel.out
   ${base_db_dir}/sbin/start-cli.sh -e " delete from "$LINE".* where time=2034-01-24T18:53:22.000+08:00" >> sel.out
   ${base_db_dir}/sbin/start-cli.sh -e " delete from "$LINE".* where time=2024-05-24T18:00:05.000+08:00" >> sel.out
   ${base_db_dir}/sbin/start-cli.sh -e " select s_0 from "$LINE" where time=2024-05-24T18:00:05.000+08:00 or time=2034-01-24T18:53:22.000+08:00" >> sel.out
   echo $LINE

done < ${out_dev}
}
del2rows

delete 2条记录成功，delete后查询这2条记录是否存在，不存在，表现正确。
脚本执行完，killall -u liuzhen

备份data到/data/liuzhen_test/0310_master_f133c92/data_1_del-2rows-when-inner-comp-kill-u
备份日志/data/liuzhen_test/0310_master_f133c92/logs.1_del-2rows-when-inner-comp-kill-u

3. 启动iotdb
启动成功后，再次执行查询2条已经delete的记录，不存在，通过。
启动q.sh，查询每个设备的count，发现有20设备存在少数据的现象。

Issue Links

• links to: GitHub Pull Request #5248
• links to: GitHub Pull Request #5249