Project

General

Profile

Actions
Copy link

Bug #13081

closed

data on rbd image get corrupted when pool quota is smaller than the size of the rbd image

Added by runsisi hust over 8 years ago. Updated almost 8 years ago.

Status:
Resolved
Priority:
Normal
Assignee:
Ilya Dryomov
Category:
libceph
Target version:
-
% Done:

0%

Source:
other
Tags:
Backport:
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Crash signature (v1):
Crash signature (v2):

Description

1. linux kernel is 4.1.1

[root@c8 ceph]# uname -r
4.1.1

2. ceph version of the cluster is 0.94.1

[root@ceph1 ~]# ceph -v
ceph version 0.94.1 (e4bfad3a3c51054df7e537a724c8d0bf9be972ff)

3. create a pool p500m with a quota of 500M-bytes:

[root@c8 ceph]# ceph osd pool create p500m 256
[root@c8 ceph]# ceph osd pool set-quota p500m max_bytes 500000000

4. create an image i1000m with a size of 1000M-bytes:

[root@c8 ceph]# rbd create --image-format 2 --size 1000 p500m/i1000m

5. map, mkfs and then mount

[root@c8 ceph]# rbd map p500m/i1000m
[root@c8 ceph]# mkfs -t ext4 /dev/rbd2
[root@c8 ceph]# mkdir -p /p500m/i1000m
[root@c8 ceph]# mount /dev/rbd2 /p500m/i1000m/

6. 800m.dat is our test data

[root@c8 ceph]# ll -h 800m.dat 
-rw-r--r--. 1 root root 800M Sep 15 04:55 800m.dat
[root@c8 ceph]# md5sum 800m.dat 
4d6c96c9be3426e12caec272da25aba1  800m.dat

7. copy test data to image

[root@c8 ceph]# cp 800m.dat /p500m/i1000m/

8. re-mount

[root@c8 ceph]# umount /dev/rbd2
[root@c8 ceph]# mount /dev/rbd2 /p500m/i1000m/

9. data is corrupted

[root@c8 ceph]# md5sum /p500m/i1000m/800m.dat
f87b61a45cd3451e512965f2ec06a226  /p500m/i1000m/800m.dat

BTW, if no quota is set on the pool, then everything goes well.

Related issues 1 (0 open — 1 closed)

Related to Linux kernel client - Feature #9779: libceph: sync up with objecterResolvedIlya Dryomov10/14/2014

Actions
Actions
Copy link
 #1

Updated by Josh Durgin over 8 years ago

  • Project changed from Ceph to Linux kernel client
  • Category set to libceph

This is due to http://tracker.ceph.com/issues/12018 - fixed on the osd and userspace client. Moving this to the kernel client section to track the fix for the kernel client.

Actions
Copy link
 #2

Updated by Ilya Dryomov almost 8 years ago

  • Assignee set to Ilya Dryomov
Actions
Copy link
 #3

Updated by Ilya Dryomov almost 8 years ago

  • Status changed from New to In Progress
Actions
Copy link
 #4

Updated by Ilya Dryomov almost 8 years ago

Actions
Copy link
 #5

Updated by Ilya Dryomov almost 8 years ago

  • Status changed from In Progress to Resolved

Done in 4.7. The umount in step 8 blocks until the quota is increased - there is nothing else we (the block device) can do. This is no different from an under-provisioned thin-pool.

libceph: FULL or reached pool quota
in the dmesg.

Actions
Copy link

Also available in: Atom PDF