What is the role of open-behind？

[jifengzhou]

After reading the open-behind xlator source code, I don’t understand which scenarios can be optimized after open-behind is turned on.
Hope someone can help explain, thanks

[pranithk]

After reading the open-behind xlator source code, I don’t understand which scenarios can be optimized after open-behind is turned on.
Hope someone can help explain, thanks

When you have read workloads, to read a file you do open, read, both of which go over network. With open-behind on, only read will go over network. For write workloads opens are sent when write happens. There are more cases to handle unlink/rename etc I guess. But this is the gist.

[jifengzhou]

@pranithk Thank you. Can I interpret anonymous files as temporary files stored in memory?

[pranithk]

@pranithk Thank you. Can I interpret anonymous files as temporary files stored in memory?

No, there are no anonymous files. Only anonymous fds. Anonymous fds are fds that didn't open the files/directories but they can do operations on them as if they opened them using the fd.

[jifengzhou]

@pranithk I don’t know when will use anonymous fd, please help give me an example, when will use anonymous fd, thank you

[amarts]

The primary reason we did anonymous fd was nfs. NFS protocol doens't have open, instead works on a filehandle which it creates in lookup. Hence we needed to operate read / write on gfid/filehandle.

As anonymous fd was available, we implemented open-behind using the same logic.

[jifengzhou]

@pranithk @amarts Got it, thank you very much