[LU-825] lfs getstrip fails on a file when parent dir cannot be read.

• Status: Closed
• Project: Lustre
• Affects Version/s: Lustre 2.0.0, Lustre 2.1.0, Lustre 2.2.0, Lustre 1.8.7, Lustre 1.8.6
• Type: Bug
• Priority: Minor
• Reporter: Richard Henwood (Inactive)
• Assignee: Richard Henwood (Inactive)
• Resolution: Won't Fix
• Severity: 3

Description

to reproduce:

[build@rhel6_21 ~]$ mkdir -p /mnt/lustre/foo/bar
[build@rhel6_21 ~]$ touch /mnt/lustre/foo/bar/baz
[build@rhel6_21 ~]$ chmod ugo-r /mnt/lustre/foo/bar
[build@rhel6_21 ~]$ lfs getstripe /mnt/lustre/foo/bar/baz
error: getstripe failed for /mnt/lustre/foo/bar/baz.
[build@rhel6_21 ~]$ 
[build@rhel6_21 ~]$ 
[build@rhel6_21 ~]$ 
[build@rhel6_21 ~]$ # check that the file is really ok:
[build@rhel6_21 ~]$ stat /mnt/lustre/foo/bar/baz
  File: `/mnt/lustre/foo/bar/baz'
  Size: 0         	Blocks: 0          IO Block: 2097152 regular empty file
Device: 2c54f966h/743766374d	Inode: 144115205255725059  Links: 1
Access: (0664/-rw-rw-r--)  Uid: (  500/   build)   Gid: (  500/   build)
Access: 2011-11-04 20:50:11.000000000 -0400
Modify: 2011-11-04 20:50:11.000000000 -0400
Change: 2011-11-04 20:50:11.000000000 -0400

This is effect is most likely caused by the implementation of getstripe within lfs. When getstripe is called with against a file, getstripe opens the parent directory of the file in question.

Comments

[Richard Henwood (Inactive)]

After discussion with Andreas: the current implementation avoids instantiating inodes on the client for every single file that is found by lfs getstripe -r. Otherwise, if each file is opened to do an ioctl to get the layout, this will increase the RPC traffic to the client dramatically

[Richard Henwood (Inactive)]

Will this implementation (opening the parent dir) play nicely with DNE - when it lands?

[Richard Henwood (Inactive)]

I haven't had any interest in this in 3 years, so I'm closing it.