[IOTDB-1903] The file disk space usage increases abnormally after IoTDB restarts

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.12.2
• Fix Version/s: 0.12.3
• Labels: pull-request-available
• Sprint: 2021-10-DragonGate
• Assignee: Liuxuxin
• Reporter: Liuxuxin
• Created: 28/Oct/21 12:40
• Updated: 05/Nov/21 20:43
• Resolved: 01/Nov/21 13:43

Description

1. Description
   If the IoTDB is shut down when cross compaction when IoTDB is restarted, there will be abnormal redundant data in TsFile, but the abnormal data cannot be seen from the client.
2. Cause
   1. The target file of cross-space compaction is deleted when sgp recover, the cross-space compaction recover can never find the target file.
   2. The MergeLogAnalyzer searches for a log that is never written to the log file.
   3. The resume cross-space compaction process gets the schema from merge resources, but the schema list in merge resources is always empty when performing recovery.
   4. The recovery of cross-space compaction uses an InputStream to truncate the file, which is forbidden by JVM.
   5. The ChunkMetadata list in the last chunk group is lost when using a RestorableTsFileIOWriter to continue to write on the target file.

 

 

Issue Links

• links to: GitHub Pull Request #4312