[Gluster-users] mv lost some files ?

Raghavendra Gowdappa rgowdapp at redhat.com
Wed Sep 5 04:40:54 UTC 2018


On Wed, Sep 5, 2018 at 9:21 AM, yu sun <sunyu1949 at gmail.com> wrote:

> thank you , I will investigate and have a try.
>

Please get all the extended attributes of src and dst from all bricks. That
should help us to (dis)prove whether its same bug.


> BYW, I found another thing:
> 1. I found this only occur when I try to mv dir to a subdir.
> 2. I create a similar volume with default option and without quota
> enabled, i found its ok, mv seems no problem.
>
> so I think it's possibly the volume option led to this problem, but I dont
> know which options and  how to resolve this, the volume have about 25T data.
>
> Best Regards.
>
> Raghavendra Gowdappa <rgowdapp at redhat.com> 于2018年9月5日周三 上午10:50写道:
>
>>
>>
>> On Tue, Sep 4, 2018 at 5:28 PM, yu sun <sunyu1949 at gmail.com> wrote:
>>
>>> Hi all:
>>>
>>> I have a replicated volume project2 with info:
>>> Volume Name: project2 Type: Distributed-Replicate Volume ID:
>>> 60175b8e-de0e-4409-81ae-7bb5eb5cacbf Status: Started Snapshot Count: 0
>>> Number of Bricks: 84 x 2 = 168 Transport-type: tcp Bricks: Brick1:
>>> node20:/data2/bricks/project2 Brick2: node21:/data2/bricks/project2 Brick3:
>>> node22:/data2/bricks/project2 Brick4: node23:/data2/bricks/project2 Brick5:
>>> node24:/data2/bricks/project2 Brick6: node25:/data2/bricks/project2 Brick7:
>>> node26:/data2/bricks/project2 Brick8: node27:/data2/bricks/project2 Brick9:
>>> node28:/data2/bricks/project2 Brick10: node29:/data2/bricks/project2
>>> Brick11: node30:/data2/bricks/project2 Brick12:
>>> node31:/data2/bricks/project2 Brick13: node32:/data2/bricks/project2
>>> Brick14: node33:/data2/bricks/project2 Brick15:
>>> node20:/data3/bricks/project2 Brick16: node21:/data3/bricks/project2
>>> Brick17: node22:/data3/bricks/project2 Brick18:
>>> node23:/data3/bricks/project2 Brick19: node24:/data3/bricks/project2
>>> Brick20: node25:/data3/bricks/project2 Brick21:
>>> node26:/data3/bricks/project2 Brick22: node27:/data3/bricks/project2
>>> Brick23: node28:/data3/bricks/project2 Brick24:
>>> node29:/data3/bricks/project2 Brick25: node30:/data3/bricks/project2
>>> Brick26: node31:/data3/bricks/project2 Brick27:
>>> node32:/data3/bricks/project2 Brick28: node33:/data3/bricks/project2
>>> Brick29: node20:/data4/bricks/project2 Brick30:
>>> node21:/data4/bricks/project2 Brick31: node22:/data4/bricks/project2
>>> Brick32: node23:/data4/bricks/project2 Brick33:
>>> node24:/data4/bricks/project2 Brick34: node25:/data4/bricks/project2
>>> Brick35: node26:/data4/bricks/project2 Brick36:
>>> node27:/data4/bricks/project2 Brick37: node28:/data4/bricks/project2
>>> Brick38: node29:/data4/bricks/project2 Brick39:
>>> node30:/data4/bricks/project2 Brick40: node31:/data4/bricks/project2
>>> Brick41: node32:/data4/bricks/project2 Brick42:
>>> node33:/data4/bricks/project2 Brick43: node20:/data5/bricks/project2
>>> Brick44: node21:/data5/bricks/project2 Brick45:
>>> node22:/data5/bricks/project2 Brick46: node23:/data5/bricks/project2
>>> Brick47: node24:/data5/bricks/project2 Brick48:
>>> node25:/data5/bricks/project2 Brick49: node26:/data5/bricks/project2
>>> Brick50: node27:/data5/bricks/project2 Brick51:
>>> node28:/data5/bricks/project2 Brick52: node29:/data5/bricks/project2
>>> Brick53: node30:/data5/bricks/project2 Brick54:
>>> node31:/data5/bricks/project2 Brick55: node32:/data5/bricks/project2
>>> Brick56: node33:/data5/bricks/project2 Brick57:
>>> node20:/data6/bricks/project2 Brick58: node21:/data6/bricks/project2
>>> Brick59: node22:/data6/bricks/project2 Brick60:
>>> node23:/data6/bricks/project2 Brick61: node24:/data6/bricks/project2
>>> Brick62: node25:/data6/bricks/project2 Brick63:
>>> node26:/data6/bricks/project2 Brick64: node27:/data6/bricks/project2
>>> Brick65: node28:/data6/bricks/project2 Brick66:
>>> node29:/data6/bricks/project2 Brick67: node30:/data6/bricks/project2
>>> Brick68: node31:/data6/bricks/project2 Brick69:
>>> node32:/data6/bricks/project2 Brick70: node33:/data6/bricks/project2
>>> Brick71: node20:/data7/bricks/project2 Brick72:
>>> node21:/data7/bricks/project2 Brick73: node22:/data7/bricks/project2
>>> Brick74: node23:/data7/bricks/project2 Brick75:
>>> node24:/data7/bricks/project2 Brick76: node25:/data7/bricks/project2
>>> Brick77: node26:/data7/bricks/project2 Brick78:
>>> node27:/data7/bricks/project2 Brick79: node28:/data7/bricks/project2
>>> Brick80: node29:/data7/bricks/project2 Brick81:
>>> node30:/data7/bricks/project2 Brick82: node31:/data7/bricks/project2
>>> Brick83: node32:/data7/bricks/project2 Brick84:
>>> node33:/data7/bricks/project2 Brick85: node20:/data8/bricks/project2
>>> Brick86: node21:/data8/bricks/project2 Brick87:
>>> node22:/data8/bricks/project2 Brick88: node23:/data8/bricks/project2
>>> Brick89: node24:/data8/bricks/project2 Brick90:
>>> node25:/data8/bricks/project2 Brick91: node26:/data8/bricks/project2
>>> Brick92: node27:/data8/bricks/project2 Brick93:
>>> node28:/data8/bricks/project2 Brick94: node29:/data8/bricks/project2
>>> Brick95: node30:/data8/bricks/project2 Brick96:
>>> node31:/data8/bricks/project2 Brick97: node32:/data8/bricks/project2
>>> Brick98: node33:/data8/bricks/project2 Brick99:
>>> node20:/data9/bricks/project2 Brick100: node21:/data9/bricks/project2
>>> Brick101: node22:/data9/bricks/project2 Brick102:
>>> node23:/data9/bricks/project2 Brick103: node24:/data9/bricks/project2
>>> Brick104: node25:/data9/bricks/project2 Brick105:
>>> node26:/data9/bricks/project2 Brick106: node27:/data9/bricks/project2
>>> Brick107: node28:/data9/bricks/project2 Brick108:
>>> node29:/data9/bricks/project2 Brick109: node30:/data9/bricks/project2
>>> Brick110: node31:/data9/bricks/project2 Brick111:
>>> node32:/data9/bricks/project2 Brick112: node33:/data9/bricks/project2
>>> Brick113: node20:/data10/bricks/project2 Brick114:
>>> node21:/data10/bricks/project2 Brick115: node22:/data10/bricks/project2
>>> Brick116: node23:/data10/bricks/project2 Brick117:
>>> node24:/data10/bricks/project2 Brick118: node25:/data10/bricks/project2
>>> Brick119: node26:/data10/bricks/project2 Brick120:
>>> node27:/data10/bricks/project2 Brick121: node28:/data10/bricks/project2
>>> Brick122: node29:/data10/bricks/project2 Brick123:
>>> node30:/data10/bricks/project2 Brick124: node31:/data10/bricks/project2
>>> Brick125: node32:/data10/bricks/project2 Brick126:
>>> node33:/data10/bricks/project2 Brick127: node20:/data11/bricks/project2
>>> Brick128: node21:/data11/bricks/project2 Brick129:
>>> node22:/data11/bricks/project2 Brick130: node23:/data11/bricks/project2
>>> Brick131: node24:/data11/bricks/project2 Brick132:
>>> node25:/data11/bricks/project2 Brick133: node26:/data11/bricks/project2
>>> Brick134: node27:/data11/bricks/project2 Brick135:
>>> node28:/data11/bricks/project2 Brick136: node29:/data11/bricks/project2
>>> Brick137: node30:/data11/bricks/project2 Brick138:
>>> node31:/data11/bricks/project2 Brick139: node32:/data11/bricks/project2
>>> Brick140: node33:/data11/bricks/project2 Brick141:
>>> node20:/data12/bricks/project2 Brick142: node21:/data12/bricks/project2
>>> Brick143: node22:/data12/bricks/project2 Brick144:
>>> node23:/data12/bricks/project2 Brick145: node24:/data12/bricks/project2
>>> Brick146: node25:/data12/bricks/project2 Brick147:
>>> node26:/data12/bricks/project2 Brick148: node27:/data12/bricks/project2
>>> Brick149: node28:/data12/bricks/project2 Brick150:
>>> node29:/data12/bricks/project2 Brick151: node30:/data12/bricks/project2
>>> Brick152: node31:/data12/bricks/project2 Brick153:
>>> node32:/data12/bricks/project2 Brick154: node33:/data12/bricks/project2
>>> Brick155: node20:/data13/bricks/project2 Brick156:
>>> node21:/data13/bricks/project2 Brick157: node22:/data13/bricks/project2
>>> Brick158: node23:/data13/bricks/project2 Brick159:
>>> node24:/data13/bricks/project2 Brick160: node25:/data13/bricks/project2
>>> Brick161: node26:/data13/bricks/project2 Brick162:
>>> node27:/data13/bricks/project2 Brick163: node28:/data13/bricks/project2
>>> Brick164: node29:/data13/bricks/project2 Brick165:
>>> node30:/data13/bricks/project2 Brick166: node31:/data13/bricks/project2
>>> Brick167: node32:/data13/bricks/project2 Brick168:
>>> node33:/data13/bricks/project2 Options Reconfigured:
>>> performance.force-readdirp: on performance.write-behind: off
>>> performance.stat-prefetch: on performance.client-io-threads: on
>>> nfs.disable: on transport.address-family: inet features.quota: on
>>> features.inode-quota: on features.quota-deem-statfs: on
>>> cluster.readdir-optimize: on cluster.lookup-optimize: on
>>> dht.force-readdirp: off client.event-threads: 10 server.event-threads: 10
>>> performance.readdir-ahead: on performance.io-cache: on
>>> performance.flush-behind: on performance.cache-size: 5GB
>>> performance.cache-max-file-size: 1MB performance.write-behind-window-size:
>>> 10MB performance.read-ahead: off network.remote-dio: enable
>>> performance.strict-o-direct: disable performance.io-thread-count: 25
>>>
>>>
>>> the volume looks ok, and I mount this volume on my client machine:
>>> mount -t glusterfs -o oom-score-adj=-999 -o direct-io-mode=disable -o
>>> use-readdirp=no node20:/project2 /mnt/project2
>>>
>>> I have a directory in /mnt/project2/, but when I mv the directory to
>>> other dirs, files in the dir lost while tree or ls the dir, some files
>>> missing, my operations is list as below:
>>>
>>
>> Looks very similar to:
>> https://bugzilla.redhat.com/show_bug.cgi?id=1118762
>> https://bugzilla.redhat.com/show_bug.cgi?id=1337394
>>
>>
>>
>>> root at ml-gpu-ser129.nmg01:/mnt/project2/371_37829$ mkdir test-dir
>>> root at ml-gpu-ser129.nmg01:/mnt/project2/371_37829$ tree
>>> .
>>> ├── face_landmarks
>>> │   └── alive
>>> │       └── logs_20180823_28
>>> │           ├── info_000000.out
>>> │           ├── info_000001.out
>>> │           ├── info_000002.out
>>> │           ├── info_000003.out
>>> │           ├── info_000004.out
>>> │           ├── info_000005.out
>>> │           ├── info_000006.out
>>> │           ├── info_000007.out
>>> │           ├── info_000008.out
>>> │           ├── info_000009.out
>>> │           ├── info_000010.out
>>> │           ├── info_000011.out
>>> │           ├── info_000012.out
>>> │           ├── info_000013.out
>>> │           ├── info_000014.out
>>> │           ├── info_000015.out
>>> │           ├── info_000016.out
>>> │           ├── info_000017.out
>>> │           ├── info_000018.out
>>> │           └── info_000019.out
>>> └── test-dir
>>>
>>> 4 directories, 20 files
>>> root at ml-gpu-ser129.nmg01:/mnt/project2/371_37829$ mv face_landmarks/
>>> test-dir/
>>> root at ml-gpu-ser129.nmg01:/mnt/project2/371_37829$ tree
>>> .
>>> └── test-dir
>>>     └── face_landmarks
>>>
>>> 2 directories, 0 files
>>> root at ml-gpu-ser129.nmg01:/mnt/project2/371_37829$ cd
>>> test-dir/face_landmarks/
>>> root at ml-gpu-ser129.nmg01:/mnt/project2/371_37829/test-dir/face_landmarks$
>>> ls
>>> root at ml-gpu-ser129.nmg01:/mnt/project2/371_37829/test-dir/face_landmarks$
>>> cd ..
>>> root at ml-gpu-ser129.nmg01:/mnt/project2/371_37829/test-dir$ mv
>>> face_landmarks/ ..
>>> root at ml-gpu-ser129.nmg01:/mnt/project2/371_37829/test-dir$ cd ..
>>> root at ml-gpu-ser129.nmg01:/mnt/project2/371_37829$ tree
>>> .
>>> ├── face_landmarks
>>> │   └── alive
>>> └── test-dir
>>>
>>> 3 directories, 0 files
>>> root at ml-gpu-ser129.nmg01:/mnt/project2/371_37829$
>>>
>>> I think i make some mistakes with volume option, buti i am not sure, so
>>> how can i find the lost files?  the files seems still int the directory,
>>> because i cant remove the directory  and rm tell me "Not empty directory"
>>>
>>
>> Its likely that src and dst of mv having same gfid and that's causing the
>> issues. Can you look into both src and dst paths on all bricks? Union of
>> contents of both directories should give all the files in the src directory
>> before mv. Once found you can,
>> * keep a backup of contents of src and dst on all bricks
>> * remove trusted.gfid xattr on src and dst from all bricks
>> * remove gfid handle (.glusterfs/<first two characters of gfid>/<second
>> set of two characters of gfid>/<gfid> on each brick)
>> * disable readdirplus in entire stack (maybe you can use a tmp mount for
>> this) [1]
>> * stat src and dst on a mount point with readdirplus disabled.
>> * Now you'll see two directories src and dst on mountpoint. You  can copy
>> the contents of both into a new directory
>>
>> [1] https://lists.gluster.org/pipermail/gluster-users/2017-
>> March/030148.html
>>
>>
>>>
>>> Any suggestions is appreciated.
>>> Many Thanks
>>>
>>> Best regards
>>> Yu
>>>
>>> _______________________________________________
>>> Gluster-users mailing list
>>> Gluster-users at gluster.org
>>> https://lists.gluster.org/mailman/listinfo/gluster-users
>>>
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20180905/c6deb4ee/attachment.html>


More information about the Gluster-users mailing list