[LU-15265] e2fsck test reports "Free blocks count wrong for group #705232 (33791, counted=32768)"

• Type: Bug
• Resolution: Cannot Reproduce
• Priority: Blocker
• Affects Version/s: Lustre 2.15.0

Environment

Configuration:

20 x Lustre 2.15 clients

4 X LNet routers

Lustre client:

2.14.55_81_gc26b347

NEO build: 6.0-65-cm-21.10.24-g2c588a6

Lustre server:

2.14.55_81_gc26b347

Model SSUs

1xE1000D 1xE1000F

4 x OSS nodes with 2 x HDD OSTs and 2 x flash OSTs

• Severity: 3
• Assignee: Artem Blagodarenko (Inactive)
• Reporter: Prasannakumar Nagasubramani

Description

e2fsck test reports "Free blocks count wrong for group #705232 (33791, counted=32768)" errors.

Pass 5: Checking group summary information
Block bitmap differences:  -23109042175
Fix? no
Free blocks count wrong for group #705232 (33791, counted=32768).
Fix? no
Free blocks count wrong for group #705252 (33784, counted=32768).
Fix? no
Free blocks count wrong for group #705279 (32762, counted=32767).
Fix? no
Free blocks count wrong (98197275837, counted=98197372106).
Fix? no
Inode bitmap differences:  -2842413
Fix? no
Pass 5: Memory used: 202116k/222912k (97836k/104281k), time: 232.98/24.04/ 1.14
Pass 5: I/O read: 4805MB, write: 0MB, rate: 20.62MB/s 

Will upload the e2image and other e2fsck logs to FTP.

Issue Links

• is related to: LU-15238 lfsck crashes MDT LDISKFS-fs error (device md65): ldiskfs_xattr_inode_iget:407: comm lfsck: EA inode 2047917093 does not have LDISKFS_EA_INODE_FL flag

Comments

[Colin Faber]

Closing, Cannot reproduce.

Thanks folks!

 

-cf

 

[Artem Blagodarenko (Inactive)]

>I suppose we can resolve it and re-open later, if necessary. artem_blagodarenko, what do you think?
Agree.
>Ok, Thanks spitzcor, I'll close this (pending artem_blagodarenko's assessment), if you see it again, please re-open.
cfaber , I have no objections

[Colin Faber]

Ok, Thanks spitzcor, I'll close this (pending artem_blagodarenko's assessment), if you see it again, please re-open.

-cf

[Cory Spitz]

cfaber, no, we don't think it was associated with LU-15404, but we also can no longer reproduce it. I suppose we can resolve it and re-open later, if necessary. artem_blagodarenko, what do you think?
Thanks for following up, Colin.

[Colin Faber]

spitzcor,

Were you guys able to determine this?

-cf

[Cory Spitz]

We need to check into the possibility that this issue is a side effect of LU-15404.

[Andreas Dilger]

Artem, could you please take a look.

[Prasannakumar Nagasubramani]

e2image from the node uploaded to FTP.
/uploads/LU15265/kjcf05n05_image_out

[Prasannakumar Nagasubramani]

e2fsck logs, kern and messages are uploaded to FTP.
ftp.whamcloud.com: /uploads/LU15265

I hope /var/log/messages and /var/log/kern is pretty enough to know what happened prior to the e2fsck test. If not please let me know what else is needed.

[Andreas Dilger]

The free block count in the group descriptors is likely to be updated if there are any repairs to the filesystem, which is not shown here, so there is no way to know if this information is useful or not. The free block count in the superblock is not kept up to date while the filesystem is mounted, so this is a very common situation. Also, there is absolutely no information here about what happened to the filesystem before e2fsck was run, so there isn't really anything useful that can be done here.