[LU-18102] sanity test_27J: FAIL: /mnt/lustre/d27J.sanity/f27J.sanity: read should fail

• Type: Bug
• Resolution: Fixed
• Priority: Minor
• Fix Version/s: Lustre 2.16.0
• Affects Version/s: Lustre 2.16.0
• Environment: Ubuntu 24.04 client ; SLES 15 SP6 client
• Severity: 3
• Assignee: Zhenyu Xu
• Reporter: Maloo

Description

This issue was created by maloo for jianyu <[e-mail]>

This issue relates to the following test suite run: https://testing.whamcloud.com/test_sets/fd1502a3-be70-40ce-b2d1-c2490a7980fd

test_27J failed with the following error:

lfs setstripe: setstripe error for '/mnt/lustre/d27J.sanity/f27J.sanity': stripe already set
lfs setstripe: setstripe error for '/mnt/lustre/d27J.sanity/f27J.sanity2': stripe already set
 sanity test_27J: @@@@@@ FAIL: /mnt/lustre/d27J.sanity/f27J.sanity: read should fail 

Test session details:
clients: https://build.whamcloud.com/job/lustre-master/4558 - 6.8.0-35-generic
servers: https://build.whamcloud.com/job/lustre-master/4558 - 5.14.0-427.24.1_lustre.el9.x86_64

<<Please provide additional information about the failure here>>

VVVVVVV DO NOT REMOVE LINES BELOW, Added by Maloo for auto-association VVVVVVV
sanity test_27J - /mnt/lustre/d27J.sanity/f27J.sanity: read should fail

Issue Links

• duplicates: LU-16101 sanity test_27J: read should fail
• is related to: LU-18070 sanity test_103a: FAIL: run_acl_subtest 'misc' failed

Comments

[Peter Jones]

Merged for 2.16

[Gerrit Updater]

"Oleg Drokin <[e-mail]>" merged in patch https://review.whamcloud.com/c/fs/lustre-release/+/55977/
Subject: LU-18102 tests: skip read/write in sanity/27J for some kernels
Project: fs/lustre-release
Branch: master
Current Patch Set:
Commit: adb3d20f3c8d3a720b293d3d5977a1849a9e0cb7

[Qian Yingjin]

It will not affect PCC-RO as the file size will not be changed for readonly PCC.

 

For PCC-RW, the file data will save into the copy on HSM backend with HSM released state.

In this case, a client other than the exclusive caching the file with PCC-RO mode may see the 0 byte for the Lustre file.

However, it will detect that file is in HSM released state in the higher call patch (generic_file_read|write), it will restore the file and then restart the I/O.

 

So it will not affect PCC-RW also, I think.

[Bruno Faccini]

Hello Andreas,
It has been quite a long time that I have been involved with this, but I seem to remember that the DAOS binding thru foreign layout is done at open time not read.
I will try to fully confirm if I can get some spare time to do some testing.

[Andreas Dilger]

bfaccini-nvda, any idea on whether/how this would affect DAOS usage of Foreign file layouts, or are they trapped before the read gets to the Lustre file?

qian_wc, will this affect PCC usage of Foreign file layouts?

[Zhenyu Xu]

It seems like a bug in Lustre if the file size is 0 but there is data in the file?

I guess not, a file created with foreign layout in this case is sized 0 and contains no data, and it would be reported ENODATA in Lustre I/O path, while kernel bypasses the readpage() so the test script would report R/W success while it expects them to fail the I/O.

[Andreas Dilger]

The kernel introduce this issue in fiemap_read() to skip calling readpage() when the file size is 0

It seems like a bug in Lustre if the file size is 0 but there is data in the file?

[Andreas Dilger]

"Zhenyu Xu <[e-mail]>" uploaded a new patch: https://review.whamcloud.com/c/fs/lustre-release/+/55977
Subject: LU-18012 tests: skip read/write in sanity/27J for some kernels
Project: fs/lustre-release
Branch: master
Current Patch Set: 1
Commit: 496e8bd36cbf26b1c694d32c719c7636f1975693