[IOTDB-6075] Pipe: File resource races when different tsfile load operations concurrently modify the same tsfile at receiver

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Labels: pull-request-available
• Assignee: Unassigned
• Reporter: Steve Yurong Su
• Created: 20/Jul/23 10:16
• Updated: 21/Jul/23 03:51

Description

How to reproduce:

1. sender's connection timeout at sender < load operation cost time at receiver
2. a sender's load tsfile request is timeout, but the request's execution can not be terminated when timeout
3. sender will retry to send a load tsfile request, which will load the same tsfile that cause the timeout before
4. serveral load tsfile requests are concurrently executing at the receiver
5. the load operation performed earlier deletes files that need to be loaded
6. the load operation performed later (a retry) wants to read the tsfile, but the tsfile was deleted, which may cause exception like: 
   writing file sequence-root.bw-6-2735-1686907665964-1-8-8.tsfile is not available

Issue Links

• links to: GitHub Pull Request #10629
• links to: GitHub Pull Request #10633