sst not found during compaction

[zouxiang1993]

Describe this problem

I create a partition table and write points to it. Then I execute a query and it fails with the response:

{"code":500,"message":"Failed to handle request, err:Failed to execute interpreter, query:\nselect count(*) from perflog\n, err:Failed to execute select, err:Failed to execute logical plan, err:Failed to collect record batch stream, err:Stream error, msg:convert from arrow record batch, err:External error: Stream error, msg:poll read response failed, err:Failed to query from table in server, table_ident:TableIdentifier { catalog: \"ceresdb\", schema: \"public\", table: \"__perflog_95\" }, code:500, msg:record batch failed. Caused by: Stream error, msg:read record batch, err:Failed to read data from the sub iterator, err:PullRecordBatch { source: DecodeRecordBatch { source: ParquetError { source: General(\"Failed to fetch ranges from object store, err:Object at location /data/ceresdb/store/0/96/182.sst not found: No such file or directory (os error 2)\"), backtrace:"}

I retry the query after a few seconds and it success.
There is no useful information in the server log.

It seems that the sst has been removed during compaction but the query gets the expired version.

Server version

CeresDB Server
CeresDB version: 1.0.0
Git branch: main
Git commit: 95ea870
Build time: 2023-03-06T03:30:09.845055021Z
Rustc version: 1.69.0-nightly

Steps to reproduce

None

Expected behavior

No response

Additional Information

No response

[chunshao90]

#699 has fixed this issue.

[ShiKaiWi]

@zouxiang1993 If you want to use the latest ceresdb server, the nightly image may be a good choice: https://github.com/CeresDB/ceresdb/pkgs/container/ceresdb-server