[LU-8039] Running lfs hsm commands results in "No data available"

• Type: Bug
• Resolution: Unresolved
• Priority: Major
• Affects Version/s: Lustre 2.8.0
• Labels: HSM
• Severity: 3
• Project: HSM
• Assignee: WC Triage
• Reporter: Tatsushi Takamura

Description

When running multiple lfs hsm commands in parallel, a file becomes inaccessible.

Here is a reproducer:

# cd /lustre
# echo aaa > file1
# cat file1
aaa
# while true; do lfs hsm_archive file1;  lfs hsm_release file1;  lfs hsm_restore file1;  lfs hsm_remove file1; done
# while true; do lfs hsm_archive file1;  lfs hsm_release file1;  lfs hsm_restore file1;  lfs hsm_remove file1; done (run on another terminal)
<wait for a while>
# lfs hsm_state file1
file1: (0x00000005) released exists, archive_id:1
# cat file1
cat: file1: No data available

From the result of lfs hsm_state, it appears that the file data which has not been archived has been released.