[LU-7701] Lost file/object?

• Status: Resolved
• Project: Lustre
• Affects Version/s: Lustre 2.5.3
• Type: Bug
• Priority: Major
• Reporter: Mahmoud Hanafi
• Assignee: Niu Yawei (Inactive)
• Resolution: Cannot Reproduce
• Severity: 3

Description

A user wrote a very large single stripe file which almost filed an OST.
We copied the file re-striping and renaming it.
Removed the original file
And renamed the file back to the original name.

We suspect that the original object didn't get deleted because the ost is still %96 full. I fund the object lma and fid.
fid: [0x20006132a:0x8256:0x0]
lma: [0x100000000:0xb9bd1:0x0]

The fid doesn't match any file
pfe21 ~ # lfs fid2path /nobackupp6 [0x20006132a:0x8256:0x0]
fid2path: error on FID [0x20006132a:0x8256:0x0]: No such file or directory

I unmounted the OST ran fsck and remounted it. It didn't help.

How do we locate the file for the givin fid?

nbp6-oss7 ~ # df
Filesystem            1K-blocks       Used  Available Use% Mounted on
/dev/sda8             240505984    8445108  219909760   4% /
tmpfs                  65936716          0   65936716   0% /dev/shm
/dev/sda7                141845      42733      91788  32% /boot
/dev/mapper/nbp6-ost46
                     7572561480 5640464516 1932096964  75% /mnt/lustre/nbp6-ost46
/dev/mapper/nbp6-ost54
                     7572561480 5745597312 1826964168  76% /mnt/lustre/nbp6-ost54
/dev/mapper/nbp6-ost94
                     7572561480 5672980392 1899581088  75% /mnt/lustre/nbp6-ost94
/dev/mapper/nbp6-ost70
                     7572561480 5681848384 1890713096  76% /mnt/lustre/nbp6-ost70
/dev/mapper/nbp6-ost118
                     7572561480 5782306340 1790255140  77% /mnt/lustre/nbp6-ost118
/dev/mapper/nbp6-ost38
                     7572561480 5575430256 1997131224  74% /mnt/lustre/nbp6-ost38
/dev/mapper/nbp6-ost30
                     7572561480 5567585580 2004975900  74% /mnt/lustre/nbp6-ost30
/dev/mapper/nbp6-ost86
                     7572561480 5835984468 1736577012  78% /mnt/lustre/nbp6-ost86
/dev/mapper/nbp6-ost6
                     7572561480 5690988184 1881573296  76% /mnt/lustre/nbp6-ost6
/dev/mapper/nbp6-ost110
                     7572561480 5687466608 1885094872  76% /mnt/lustre/nbp6-ost110
/dev/mapper/nbp6-ost102
                     7572561480 5893025996 1679535484  78% /mnt/lustre/nbp6-ost102
/dev/mapper/nbp6-ost78
                     7572561480 5699376592 1873184888  76% /mnt/lustre/nbp6-ost78
/dev/mapper/nbp6-ost22
                     7572561480 5678322400 1894239080  75% /mnt/lustre/nbp6-ost22
/dev/mapper/nbp6-ost62
                     7572561480 5629283036 1943278444  75% /mnt/lustre/nbp6-ost62
/dev/mapper/nbp6-ost14
                     7572561480 7221376872  351184608  96% /mnt/lustre/nbp6-ost14

Comments

[Peter Jones]

Niu

Could you please advise?

Thanks

Peter

[Niu Yawei (Inactive)]

The fid looks like from a MDT object, but the lma looks like from OST object, could you verify how did you get the fid and the lma?

If the lma (lma: [0x100000000:0xb9bd1:0x0]) is from the OST object, then it represents the self-fid of the OST object, you can check the object debugfs:

debugfs -c -R "stat /O/0/d$((760785 % 32))/760785" $ostdev". (760785 is the decimal of 0xb9bd1, and $ostdev is the OST device where the object is located in).

[Mahmoud Hanafi]

I read all the inode on the OST and get the lma and fid from the ost inode info.

But it looks like the object was cleaned up at some point. It no longer exists. I am not sure why it took so long.

You may close this case.

[Peter Jones]

ok - thanks Mahmoud