[Gluster-users] Brick missing trusted.glusterfs.dht xattr

Sunny Kumar sunkumar at redhat.com
Mon Jul 29 05:56:59 UTC 2019


HI Matthew,

Can you share geo-rep logs and one more log file
(changes-<brick-path>.log) it will help to pinpoint actual reason
behind failure.

/sunny

On Mon, Jul 29, 2019 at 9:13 AM Nithya Balachandran <nbalacha at redhat.com> wrote:
>
>
>
> On Sat, 27 Jul 2019 at 02:31, Matthew Benstead <matthewb at uvic.ca> wrote:
>>
>> Ok thank-you for explaining everything - that makes sense.
>>
>> Currently the brick file systems are pretty evenly distributed so I probably won't run the fix-layout right now.
>>
>> Would this state have any impact on geo-replication? I'm trying to geo-replicate this volume, but am getting a weird error: "Changelog register failed error=[Errno 21] Is a directory"
>
>
> It should not. Sunny, can you comment on this?
>
> Regards,
> Nithya
>>
>>
>> I assume this is related to something else, but I wasn't sure.
>>
>> Thanks,
>>  -Matthew
>>
>> --
>> Matthew Benstead
>> System Administrator
>> Pacific Climate Impacts Consortium
>> University of Victoria, UH1
>> PO Box 1800, STN CSC
>> Victoria, BC, V8W 2Y2
>> Phone: +1-250-721-8432
>> Email: matthewb at uvic.ca
>>
>> On 7/26/19 12:02 AM, Nithya Balachandran wrote:
>>
>>
>>
>> On Fri, 26 Jul 2019 at 01:56, Matthew Benstead <matthewb at uvic.ca> wrote:
>>>
>>> Hi Nithya,
>>>
>>> Hmm... I don't remember if I did, but based on what I'm seeing it sounds like I probably didn't run rebalance or fix-layout.
>>>
>>> It looks like folders that haven't had any new files created have a dht of 0, while other folders have non-zero values.
>>>
>>> [root at gluster07 ~]# getfattr --absolute-names -m . -d -e hex /mnt/raid6-storage/storage/ | grep dht
>>> [root at gluster07 ~]# getfattr --absolute-names -m . -d -e hex /mnt/raid6-storage/storage/home | grep dht
>>> trusted.glusterfs.dht=0x00000000000000000000000000000000
>>> [root at gluster07 ~]# getfattr --absolute-names -m . -d -e hex /mnt/raid6-storage/storage/home/matthewb | grep dht
>>> trusted.glusterfs.dht=0x00000001000000004924921a6db6dbc7
>>>
>>> If I just run the fix-layout command will it re-create all of the dht values or just the missing ones?
>>
>>
>> A fix-layout will recalculate the layouts entirely so files all the values will change. No files will be moved.
>> A rebalance will recalculate the layouts like the fix-layout but will also move files to their new locations based on the new layout ranges. This could take a lot of time depending on the number of files/directories on the volume. If you do this, I would recommend that you turn off lookup-optimize until the rebalance is over.
>>
>>>
>>> Since the brick is already fairly size balanced could I get away with running fix-layout but not rebalance? Or would the new dht layout mean slower accesses since the files may be expected on different bricks?
>>
>>
>> The first access for a file will be slower. The next one will be faster as the location will be cached in the client's in-memory structures.
>> You may not need to run either a fix-layout or a rebalance if new file creations will be in directories created after the add-brick. Gluster will automatically include all 7 bricks for those directories.
>>
>> Regards,
>> Nithya
>>
>>>
>>> Thanks,
>>>  -Matthew
>>>
>>> --
>>> Matthew Benstead
>>> System Administrator
>>> Pacific Climate Impacts Consortium
>>> University of Victoria, UH1
>>> PO Box 1800, STN CSC
>>> Victoria, BC, V8W 2Y2
>>> Phone: +1-250-721-8432
>>> Email: matthewb at uvic.ca
>>>
>>> On 7/24/19 9:30 PM, Nithya Balachandran wrote:
>>>
>>>
>>>
>>> On Wed, 24 Jul 2019 at 22:12, Matthew Benstead <matthewb at uvic.ca> wrote:
>>>>
>>>> So looking more closely at the trusted.glusterfs.dht attributes from the bricks it looks like they cover the entire range... and there is no range left for gluster07.
>>>>
>>>> The first 6 bricks range from 0x00000000 to 0xffffffff - so... is there a way to re-calculate what the dht values should be? Each of the bricks should have a gap
>>>>
>>>> Gluster05 00000000 -> 2aaaaaa9
>>>> Gluster06 2aaaaaaa -> 55555553
>>>> Gluster01 55555554 -> 7ffffffd
>>>> Gluster02 7ffffffe -> aaaaaaa7
>>>> Gluster03 aaaaaaa8 -> d5555551
>>>> Gluster04 d5555552 -> ffffffff
>>>> Gluster07 None
>>>>
>>>> If we split the range into 7 servers that would be a gap of about 0x24924924 for each server.
>>>>
>>>> Now in terms of the gluster07 brick, about 2 years ago the RAID array the brick was stored on became corrupted. I ran the remove-brick force command, then provisioned a new server, ran the add-brick command and then restored the missing files from backup by copying them back to the main gluster mount (not the brick).
>>>>
>>>
>>> Did you run a rebalance after performing the add-brick? Without a rebalance/fix-layout , the layout for existing directories on the volume will not  be updated to use the new brick as well.
>>>
>>> That the layout does not include the new brick in the root dir is in itself is not a problem. Do you create a lot of files directly in the root of the volume? If yes, you might want to run a rebalance. Otherwise, if you mostly create files in newly added directories, you can probably ignore this. You can check the layout for directories on the volume and see if they incorporate the brick7.
>>>
>>> I would expect a lookup on the root to have set an xattr on the brick with an empty layout range . The fact that the xattr does not exist at all on the brick is what I am looking into.
>>>
>>>
>>>> It looks like prior to that event this was the layout - which would make sense given the equal size of the 7 bricks:
>>>>
>>>> gluster02.pcic.uvic.ca | SUCCESS | rc=0 >>
>>>> # file: /mnt/raid6-storage/storage
>>>> trusted.glusterfs.dht=0x000000010000000048bfff206d1ffe5f
>>>>
>>>> gluster05.pcic.uvic.ca | SUCCESS | rc=0 >>
>>>> # file: /mnt/raid6-storage/storage
>>>> trusted.glusterfs.dht=0x0000000100000000b5dffce0da3ffc1f
>>>>
>>>> gluster04.pcic.uvic.ca | SUCCESS | rc=0 >>
>>>> # file: /mnt/raid6-storage/storage
>>>> trusted.glusterfs.dht=0x0000000100000000917ffda0b5dffcdf
>>>>
>>>> gluster03.pcic.uvic.ca | SUCCESS | rc=0 >>
>>>> # file: /mnt/raid6-storage/storage
>>>> trusted.glusterfs.dht=0x00000001000000006d1ffe60917ffd9f
>>>>
>>>> gluster01.pcic.uvic.ca | SUCCESS | rc=0 >>
>>>> # file: /mnt/raid6-storage/storage
>>>> trusted.glusterfs.dht=0x0000000100000000245fffe048bfff1f
>>>>
>>>> gluster07.pcic.uvic.ca | SUCCESS | rc=0 >>
>>>> # file: /mnt/raid6-storage/storage
>>>> trusted.glusterfs.dht=0x000000010000000000000000245fffdf
>>>>
>>>> gluster06.pcic.uvic.ca | SUCCESS | rc=0 >>
>>>> # file: /mnt/raid6-storage/storage
>>>> trusted.glusterfs.dht=0x0000000100000000da3ffc20ffffffff
>>>>
>>>> Which yields the following:
>>>>
>>>> 00000000 -> 245fffdf    Gluster07
>>>> 245fffe0 -> 48bfff1f    Gluster01
>>>> 48bfff20 -> 6d1ffe5f    Gluster02
>>>> 6d1ffe60 -> 917ffd9f    Gluster03
>>>> 917ffda0 -> b5dffcdf    Gluster04
>>>> b5dffce0 -> da3ffc1f    Gluster05
>>>> da3ffc20 -> ffffffff    Gluster06
>>>>
>>>> Is there some way to get back to this?
>>>>
>>>> Thanks,
>>>>  -Matthew
>>>>
>>>> --
>>>> Matthew Benstead
>>>> System Administrator
>>>> Pacific Climate Impacts Consortium
>>>> University of Victoria, UH1
>>>> PO Box 1800, STN CSC
>>>> Victoria, BC, V8W 2Y2
>>>> Phone: +1-250-721-8432
>>>> Email: matthewb at uvic.ca
>>>>
>>>> On 7/18/19 7:20 AM, Matthew Benstead wrote:
>>>>
>>>> Hi Nithya,
>>>>
>>>> No - it was added about a year and a half ago. I have tried re-mounting the volume on the server, but it didn't add the attr:
>>>>
>>>> [root at gluster07 ~]# umount /storage/
>>>> [root at gluster07 ~]# cat /etc/fstab | grep "/storage"
>>>> 10.0.231.56:/storage /storage glusterfs defaults,log-level=WARNING,backupvolfile-server=10.0.231.51 0 0
>>>> [root at gluster07 ~]# mount /storage/
>>>> [root at gluster07 ~]# df -h /storage/
>>>> Filesystem            Size  Used Avail Use% Mounted on
>>>> 10.0.231.56:/storage  255T  194T   62T  77% /storage
>>>> [root at gluster07 ~]# getfattr --absolute-names -m . -d -e hex /mnt/raid6-storage/storage/
>>>> # file: /mnt/raid6-storage/storage/
>>>> security.selinux=0x756e636f6e66696e65645f753a6f626a6563745f723a756e6c6162656c65645f743a733000
>>>> trusted.gfid=0x00000000000000000000000000000001
>>>> trusted.glusterfs.6f95525a-94d7-4174-bac4-e1a18fe010a2.xtime=0x5d307baa00023ec0
>>>> trusted.glusterfs.quota.dirty=0x3000
>>>> trusted.glusterfs.quota.size.2=0x00001b71d5279e000000000000763e32000000000005cd53
>>>> trusted.glusterfs.volume-id=0x6f95525a94d74174bac4e1a18fe010a2
>>>>
>>>> Thanks,
>>>>  -Matthew
>>>>
>>>> On 7/17/19 10:04 PM, Nithya Balachandran wrote:
>>>>
>>>> Hi Matthew,
>>>>
>>>> Was this node/brick added to the volume recently? If yes, try mounting the volume on a fresh mount point - that should create the xattr on this as well.
>>>>
>>>> Regards,
>>>> Nithya
>>>>
>>>> On Wed, 17 Jul 2019 at 21:01, Matthew Benstead <matthewb at uvic.ca> wrote:
>>>>>
>>>>> Hello,
>>>>>
>>>>> I've just noticed one brick in my 7 node distribute volume is missing
>>>>> the trusted.glusterfs.dht xattr...? How can I fix this?
>>>>>
>>>>> I'm running glusterfs-5.3-2.el7.x86_64 on CentOS 7.
>>>>>
>>>>> All of the other nodes are fine, but gluster07 from the list below does
>>>>> not have the attribute.
>>>>>
>>>>> $ ansible -i hosts gluster-servers[0:6] ... -m shell -a "getfattr -m .
>>>>> --absolute-names -n trusted.glusterfs.dht -e hex
>>>>> /mnt/raid6-storage/storage"
>>>>> ...
>>>>> gluster05 | SUCCESS | rc=0 >>
>>>>> # file: /mnt/raid6-storage/storage
>>>>> trusted.glusterfs.dht=0x0000000100000000000000002aaaaaa9
>>>>>
>>>>> gluster03 | SUCCESS | rc=0 >>
>>>>> # file: /mnt/raid6-storage/storage
>>>>> trusted.glusterfs.dht=0x0000000100000000aaaaaaa8d5555551
>>>>>
>>>>> gluster04 | SUCCESS | rc=0 >>
>>>>> # file: /mnt/raid6-storage/storage
>>>>> trusted.glusterfs.dht=0x0000000100000000d5555552ffffffff
>>>>>
>>>>> gluster06 | SUCCESS | rc=0 >>
>>>>> # file: /mnt/raid6-storage/storage
>>>>> trusted.glusterfs.dht=0x00000001000000002aaaaaaa55555553
>>>>>
>>>>> gluster02 | SUCCESS | rc=0 >>
>>>>> # file: /mnt/raid6-storage/storage
>>>>> trusted.glusterfs.dht=0x00000001000000007ffffffeaaaaaaa7
>>>>>
>>>>> gluster07 | FAILED | rc=1 >>
>>>>> /mnt/raid6-storage/storage: trusted.glusterfs.dht: No such
>>>>> attributenon-zero return code
>>>>>
>>>>> gluster01 | SUCCESS | rc=0 >>
>>>>> # file: /mnt/raid6-storage/storage
>>>>> trusted.glusterfs.dht=0x0000000100000000555555547ffffffd
>>>>>
>>>>> Here are all of the attr's from the brick:
>>>>>
>>>>> [root at gluster07 ~]# getfattr --absolute-names -m . -d -e hex
>>>>> /mnt/raid6-storage/storage/
>>>>> # file: /mnt/raid6-storage/storage/
>>>>> security.selinux=0x756e636f6e66696e65645f753a6f626a6563745f723a756e6c6162656c65645f743a733000
>>>>> trusted.gfid=0x00000000000000000000000000000001
>>>>> trusted.glusterfs.6f95525a-94d7-4174-bac4-e1a18fe010a2.xtime=0x5d2dee800001fdf9
>>>>> trusted.glusterfs.quota.dirty=0x3000
>>>>> trusted.glusterfs.quota.size.2=0x00001b69498a1400000000000076332e000000000005cd03
>>>>> trusted.glusterfs.volume-id=0x6f95525a94d74174bac4e1a18fe010a2
>>>>>
>>>>>
>>>>> And here is the volume information:
>>>>>
>>>>> [root at gluster07 ~]# gluster volume info storage
>>>>>
>>>>> Volume Name: storage
>>>>> Type: Distribute
>>>>> Volume ID: 6f95525a-94d7-4174-bac4-e1a18fe010a2
>>>>> Status: Started
>>>>> Snapshot Count: 0
>>>>> Number of Bricks: 7
>>>>> Transport-type: tcp
>>>>> Bricks:
>>>>> Brick1: 10.0.231.50:/mnt/raid6-storage/storage
>>>>> Brick2: 10.0.231.51:/mnt/raid6-storage/storage
>>>>> Brick3: 10.0.231.52:/mnt/raid6-storage/storage
>>>>> Brick4: 10.0.231.53:/mnt/raid6-storage/storage
>>>>> Brick5: 10.0.231.54:/mnt/raid6-storage/storage
>>>>> Brick6: 10.0.231.55:/mnt/raid6-storage/storage
>>>>> Brick7: 10.0.231.56:/mnt/raid6-storage/storage
>>>>> Options Reconfigured:
>>>>> changelog.changelog: on
>>>>> features.quota-deem-statfs: on
>>>>> features.read-only: off
>>>>> features.inode-quota: on
>>>>> features.quota: on
>>>>> performance.readdir-ahead: on
>>>>> nfs.disable: on
>>>>> geo-replication.indexing: on
>>>>> geo-replication.ignore-pid-check: on
>>>>> transport.address-family: inet
>>>>>
>>>>> Thanks,
>>>>>  -Matthew
>>>>> _______________________________________________
>>>>> Gluster-users mailing list
>>>>> Gluster-users at gluster.org
>>>>> https://lists.gluster.org/mailman/listinfo/gluster-users
>>>>
>>>>
>>>>
>>>
>>


More information about the Gluster-users mailing list