[LU-7107] lfs_migrate() may fail in striped directories Created: 04/Sep/15  Updated: 08/Sep/15  Resolved: 08/Sep/15

Status: Resolved
Project: Lustre
Component/s: None
Affects Version/s: Lustre 2.8.0
Fix Version/s: None

Type: Bug Priority: Critical
Reporter: John Hammond Assignee: WC Triage
Resolution: Duplicate Votes: 0
Labels: dne2, lfs, utils

Issue Links:
Related
is related to LU-6831 The ticket for tracking all DNE2 bugs Reopened
Severity: 3
Rank (Obsolete): 9223372036854775807

 Description   

lfs migrate may fail when applied to files in a striped directory:

# export MDSCOUNT=4
# llmount.sh 
...
# cd /mnt/lustre
# lfs mkdir -i0 -c2 d0
# cd d0
# lfs setstripe -c2 f0
# lfs path2fid f0
[0x240000404:0x1:0x0]
# lfs migrate -c1 f0
lfs: f0: cannot swap layouts: Device or resource busy
cannot put lease: No locks available (37)
error: migrate: migrate file 'f0' failed
# lfs setstripe -c2 f1
# lfs path2fid f1
[0x200000404:0x3:0x0]
# lfs migrate -c1 f1
# lfs setstripe -c2 f2
# lfs path2fid f2
[0x240000404:0x2:0x0]
# lfs migrate -c1 f2
lfs: f2: cannot swap layouts: Device or resource busy
cannot put lease: No locks available (37)
error: migrate: migrate file 'f2' failed

In lfs_migrate() we must get the MDT index of the file to be migrated and use that when constructing the name of the volatile file.



 Comments   
Comment by John Hammond [ 08/Sep/15 ]

This is fixed in the latest patch set for LU-7057. See http://review.whamcloud.com/#/c/16166/.

Comment by Peter Jones [ 08/Sep/15 ]

ok - closing ticket as a dupe

Generated at Sat Feb 10 02:06:03 UTC 2024 using Jira 9.4.14#940014-sha1:734e6822bbf0d45eff9af51f82432957f73aa32c.