UnifyFS icon indicating copy to clipboard operation
UnifyFS copied to clipboard

client: attribute management for open and fsync

Open dsikich opened this issue 7 years ago • 0 comments

Avoid the conflict on file id. Current implementation generate files id based on hashcode of file name, conflict happens when two file names are hashed to the same file id.

Provide atomic update on file id. When a file is openned with O_CREAT, the client first checks if the file is created, then creates a file if it does not exist, or returns error if O_EXCL is set. These operations needs to be atomic.

During fsync, the final file size should be the largest size (typically in the checkpoint workload) among the metadata sent by all clients. Current implementation takes the last file size received as the final file size for fsync, which may not be the correct one.

dsikich avatar Dec 20 '17 01:12 dsikich