[LU-13051] ACL fail on some clients or mount points

• Type: Bug
• Resolution: Fixed
• Priority: Major
• Severity: 3
• Assignee: Peter Jones
• Reporter: Mahmoud Hanafi

Description

Sometimes set ACLs will fail on one client but will work on a different client or on a new mount on the same client.
And it can be work intermittently.

We started to see this with

Clients:
lustre2.12.2 clients and sles11sp4

Servers:
lustre2.10.8 and lustre2.12.2

For example (I changed the real user and group names)

pfe22 ~ # getfacl /nobackupp2/t12345n
getfacl: Removing leading '/' from absolute path names
# file: nobackupp2/t12345n
# owner: t12345n
# group: s0123
user::rwx
user:data123:r-x
group::---
group:s0123:r-x
mask::r-x
other::---

pfe22 ~ # su - data123 -c "ls /nobackupp2/t12345n"
ls: cannot open directory '/nobackupp2/t12345n': Permission denied

pfe22 ~ # mkdir /nobackupp2_new
pfe22 ~ # mount -t lustre 10.151.26.96@o2ib:/nbp2 /nobackupp2_new
pfe22 ~ # su - data123 -c "ls /nobackupp2_new/t12345n"
test

Issue Links

• is related to: LU-12657 sanity/103 and sanityn/25 fail with 4.18 kernel

Comments

[Peter Jones]

ok - thanks

[Mahmoud Hanafi]

Please close no longer an issue in 2.12.3

[Mahmoud Hanafi]

It could be a duplicate of LU-12657. We'll update to 2.12.3 and see if we still see the issue.

[Peter Jones]

mhanafi could this be a duplicate of LU-12657? It usually affects newer kerrnels but SUSE could well have back ported something significant...

[Mahmoud Hanafi]

I uploaded 2 debug trace one that shows acl working and one that doesn't
ftp.whamcloud.com/uploads/LU-13051/p24_worked_lookup.debug.out
ftp.whamcloud.com/uploads/LU-13051/p22_failed_lookup.debug.out