[Gluster-users] Healing issues - Solved

Mon Aug 2 15:22:02 UTC 2021

Now I would recommend you to recheck the brick layout  in such case that a server failure won't "kill" the subvolume.
For example, if your volume (replica 3 or replica 3 arbiter 1) has brick layout like:serverA:/brick/brick1serverB:/brick/brick1serverC:/brick/brick1serverD:/brick/brick2serverA:/brick/brick2serverB:/brick/brick2
the first subvolume consists of the 'brick1' (first 3 entries of the output), while 'brick2' is your second subvolume.
Each file (or a shard, when it's enabled) will be located in a single subvolume and if 2 (out of those 3) entries are on a same host -> host failure will lead to unavailability of some of the files (or shards) untill the quorum is restored ( 2 out of 3 bricks).
Best Regards,Strahil Nikolov

  On Mon, Aug 2, 2021 at 18:11, Valerio Luccio<valerio.luccio at nyu.edu> wrote:    
Thanks Strahil,

it worked like a charm.

On 7/31/21 5:21 PM, Strahil Nikolov wrote: 

 You most probably already have an /etc/fstab entry, so just recreate the LV, recreate the FS (mkfs.xfs -i size=512 /path/to/lv) and mount it.

For source brick and new brick , just use 'hydra4:/gluster1/data' .

Don't forget to test on non-prod first ;)

Best Regards,
Strahil Nikolov

В събота, 31 юли 2021 г., 19:50:31 ч. Гринуич+3, Valerio Luccio <valerio.luccio at nyu.edu> написа: 

Thanks Strahil,

I have a couple of questions. 

My /gluster1 is mounted, but no "data" folder, I suppose that will be created when I reset the brick. I see two versions of the reset-brick operations:

   gluster reset-brick <VOLNAME> <SOURCE-BRICK> start
gluster reset-brick <VOLNAME> <SOURCE-BRICK> <NEW-BRICK> commit

 I assume I should use the first version. When I specify the <SOURCE-BRICK> do I just put the number or do I have to include the "Brick" prefix ?

As for the brick layout, you hit the nail on the head. My VMs' went down and it was very painful to get everything back up to speed. I thought I had implemented the correct layout, I guess I was wrong. Which one do you suggest ?

Thanks so much,

On 7/31/21 10:54 AM, Strahil Nikolov wrote:

 Yep, you have to bring back hydra4:/gluster1/data 

You can mount again /gluster1/data (don't forget SELINUX) and then use gluster's reset-brick to rebuild that brick.

Most probably you have an entry that was saved on hydra4:/gluster1/data and the arbiter but was not pushed to the surviving brick (based on the entry in the arbiter)

Usually, I use method 2 from https://urldefense.proofpoint.com/v2/url?u=https-3A__docs.gluster.org_en_latest_Troubleshooting_gfid-2Dto-2Dpath_&d=DwIFaQ&c=slrrB7dE8n7gBJbeO0g-IQ&r=zZK0dca4HNf-XwnAN9ais1C3ncS0n2x39pF7yr-muHY&m=FMO4OcFZ-117zmwVGZre6m7D-QHTvX_5dvOJrLbalMo&s=WMkvh31SMb6jhuAJ-V5bC1pOtZX6EuVm-ARdKYwg3TE&e=  to identify the file and then you can check the status of that file on the bricks (and if necessary restore from backup). You should do that only after you have brought back hydra4:/gluster1/data to the volume.

You should also reconsider the brick layout, as in some cases a single hydra failure (whole server) will kill a subvolume and all VMs' disks on that subvolume will be unavailable.

Best Regards,

Strahil Nikolov

On Sat, Jul 31, 2021 at 2:18, Valerio Luccio

<valerio.luccio at nyu.edu> wrote:

Strahil,

did some more digging into the heal info output.

Found also the following:

   [...]
Brick hydra4:/gluster1/data                                                                                                                                                           
Status: Transport endpoint is not connected                                                                                                                                           
Number of entries: -                                                                                                                                                                  

Brick hydra3:/arbiter/2                                                                                                                                                               
<gfid:2aa223b0-77f5-441e-bc76-34c8d459eeaa> 
[...]

 So, to augment what I wrote before, the errors appear in "Brick hydra3:/gluster3/data" and "Brick hydra3:/arbiter/2", plus that "Transport endpoint is not connected" for "Brick hydra4:/gluster1/data".

I need to add that hydra4:/gluster1 is the RAID that had major hardware failure.

-- 

As a result of Coronavirus-related precautions, NYU and the Center for Brain Imaging operations will be managed remotely until further notice. 
All telephone calls and e-mail correspondence are being monitored remotely during our normal business hours of 9am-5pm, Monday through Friday. 

For MRI scanner-related emergency, please contact: Keith Sanzenbach at  keith.sanzenbach at nyu.edu  and/or Pablo Velasco at  pablo.velasco at nyu.edu 
For computer/hardware/software emergency, please contact: Valerio Luccio at  valerio.luccio at nyu.edu 
For TMS/EEG-related emergency, please contact: Chrysa Papadaniil at  chrysa at nyu.edu 
For CBI-related administrative emergency, please contact: Jennifer Mangan at  jennifer.mangan at nyu.edu 

Valerio Luccio     (212) 998-8736 
Center for Brain Imaging     4 Washington Place, Room 158 
New York University     New York, NY 10003 

 "In an open world, who needs windows or gates ?"

 -- 

| As a result of Coronavirus-related precautions, NYU and the Center for Brain Imaging operations will be managed remotely until further notice. |
| All telephone calls and e-mail correspondence are being monitored remotely during our normal business hours of 9am-5pm, Monday through Friday. |
|   |
| For MRI scanner-related emergency, please contact: Keith Sanzenbach at keith.sanzenbach at nyu.edu and/or Pablo Velasco at pablo.velasco at nyu.edu |
| For computer/hardware/software emergency, please contact: Valerio Luccio at valerio.luccio at nyu.edu |
| For TMS/EEG-related emergency, please contact: Chrysa Papadaniil at chrysa at nyu.edu |
| For CBI-related administrative emergency, please contact: Jennifer Mangan at jennifer.mangan at nyu.edu |

| Valerio Luccio |     | (212) 998-8736 |
| Center for Brain Imaging |     | 4 Washington Place, Room 158 |
| New York University |     | New York, NY 10003 |

"In an open world, who needs windows or gates ?"

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20210802/e3f8381f/attachment.html>