What are we doing here?

This blog includes a series of videos and references to help new users or enthusiasts better understand how to use open source and free technology tools. The quick links includes more information for accessing many of the tools covered along with other references to learn more for taking advantage of these tools.

Click HERE to see the full list of topics covered!

Recover a broken NextCloud

This is not intended to be a definitive guide. I am just sharing my experience after crashing my NextCloud container. 

I performed an upgrade on my "NAS" (Ubuntu server on an old laptop) by adding a second, larger external USB drive. This is not best practice, but works for me as a file backup. 

In adding the drive, I was trying to be careful, but obviously I wasn't careful enough. I changed the fstab file (this is where Linux knows what drives and filesystems to mount), to include the new drive. That generally worked fine, but I was mounting using the /dev/sda~/dev/sdb terminology for the drives. When Linux boots it checks the storage driver for what hardware and assigns it to the /dev/ based on its type (nvme will show up /dev/nvme####, SATA drives will usually show up as /dev/sdX#). What happened was because Linux assigns the /dev type at boot, when I rebooted with the new external drive the /dev assignments changed. SURPRISE!!! That then meant 2 out of 3 of my fstab assignments are wrong. They tried to mount backwards.

So that wasn't good. My root drive was using the UUID, which is better and should be used. For the other 2 external drives I changed to using the drive label that can be set when initially formatting the drives. Something like the below:

LABEL=usbbackup /media ntfs defaults 0 0

This was fine for getting everything straightened out. However, in the process of the drives mounting in the wrong order, my NextCloud and MariaDB containers got out of whack. I couldn't log in and just had an "Internal server error" message. 

The help pages mentioned that one can use maintenance mode and using the occ tools to help fix something similar. Unfortunately, the Docker install doesn't have a lot of tools, like 'sudo' to help test. Fortunately, all the data for both containers was 100% still intact as it was mounted using the volume parameter from the docker-compose file. This means, in theory, you can just nuke the containers, rebuild new containers with the same volume points, and get back in business. 

At this point, I want to stop and say that while this is possible and worked for me, be very very very careful that the data is still intact for both the database and NextCloud instances. If anything beyond the defaults in the docker-compose file was changed - i.e. additional packages or other things were added in the container - the next steps will wipe all of that. 

Steps I used:

Stop the containers:

  sudo docker stop nxtcloud nxtclouddb 

Note: Container names are defined as such in my version of the docker-compose file. Running sudo docker ps -a will show the current names of all the containers.

Backup the container data directories to somewhere safe. Leave the data intact in the original location. 

Note: If you can't get the containers restarted and need to truly nuke and pave, all the files in NextCloud are stored under the NextCloud data directory. On the container this would be /var/www/html/data/, on the local volume it would be under the <NextCloud install directory>/data/data/ on the host system using my docker-compose file.

Delete the containers:
  sudo docker rm nxtcloud nxtclouddb

Download the latest version of the NextCloud image:
 sudo docker image pull nextcloud:latest

Update the docker-compose file to specify the MariaDB version as 10.5:
 image: mariadb:10.5

This step is run because on my first go, NextCloud had issues with the latest version of MariaDB (version 10.6 I believe). The docker-compose pulled the latest image version by default.

Recreate and run the containers. From the directory of the docker-compose and data directories run:
 sudo docker-compose up -d

Once the container names appear, verify they are running correctly:
 sudo docker ps

Now open to the NextCloud browser page. In my case it asked to update the NextCloud instance. This triggered everything to get sorted out in NextCloud, and I could log in and everything was fine. Users can log in, data was intact, and things were copacetic.

The updated version of my NextCloud docker-compose file to reflect the change to use MariaDB version 10.5 is HERE.

Moral of the story, always stop running services on a server, especially the Docker containers BEFORE making a hardware update to the system. No matter how simple it may seem, put the host system in maintenance mode first.

More information on /etc/fstab configs is available here:
  https://linuxhint.com/write-edit-etc-fstab/
  https://help.ubuntu.com/community/Fstab

Hopefully this helps demonstrate how volumes can be used to keep container data persistent, along with the flexibility of working with containers. I was lucky that the upgrade was able to trigger a re-sync of the NextCloud instance and the database. Some more information about doing this manually is in the NextCloud Doc page.
  https://docs.nextcloud.com/server/latest/admin_manual/configuration_server/occ_command.html



No comments:

Post a Comment