[TiDB Usage Environment] Production Environment / Test / Poc
Test
[TiDB Version]
7.1.0
[Reproduction Path] What operations were performed when the issue occurred
Executing a large SQL query, each execution takes about 100 seconds before an error occurs. There is a panic in the logs, but no OOM. Is there a configuration issue?
The memory usage is only over 10GB, and the threshold for a single transaction is larger than this. There are no OOM logs on my tidbserverj node either.
However, currently, as long as the SQL execution exceeds 100 seconds, this error will occur… It should not be related to the configuration. I don’t know where some parameter is set.
I see that when you insert into the new table, you also formatted the time field of the source table. In that case, you should be able to query the source table’s time field with several time interval conditions and insert into the new table in batches, which will likely avoid this issue.
No problem, I found that it seems to be an issue with HAProxy. When I directly connect and write using one of the 4000 ports, there is no “lost” error.