[LU-18842] lustreapi: add a volatile function that supports layouts

• Type: Improvement
• Resolution: Unresolved
• Priority: Minor
• Labels: easy
• Severity: 3
• Assignee: Robert Read
• Reporter: Robert Read

Description

Currently lustreapi has llapi_create_volatile_param which works for basic striping, but we'll need one that also supports llapi_layout to support all the layout options.

Since this new function will likely share most of the code that is in llapi_create_volatile_param, it might make sense to create an internal
function that can handle either a param or layout, and call this from the both llapi_create_volatile_param and the new llapi_create_volatile_layout.

Issue Links

• is related to: LU-18908 hsm: Add support for a data mover separate from the copytool
• is related to: LU-18843 allow parallel rename inside a single directory
• is related to: LU-9512 Implement O_TMPFILE for Lustre

Comments

[Gerrit Updater]

"Robert Read <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/58514
Subject: LU-18842 lustreapi: create volatile file with a layout
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 2ebffb77696c6c63bf4837ae194ac7cd0392c783

[Andreas Dilger]

The intent of the llapi_generate_volatile_filename() function is to generate the filename that will create the volatile file, instead of the current code that is using 'LUSTRE_VOLATILE_HDR ":%.4X:%.4X"' directly. The details of how to create a volatile file shouldn't be spread around, but centralized in a single function.

[Robert Read]

Currently there is no way in the api  to specify a name for a volatile file so how would that generated filename be used? 

[Andreas Dilger]

It would also be useful to have an llapi_generate_volatile_filename() (or "llapi_new_volatile_filename()" or similar) function that is generating the LUSTRE_VOLATILE_HDR ":%.4X:%.4X" filename based on the MDT index, rather than hard-coding this pathname into various places in the code.

I thought I had done this in another patch somewhere, but was unable to find it.