unRAID ver 7.3.2

AppData share is deleted. The Docker tab displays, "Docker Service failed to start"

This problem was caused by migration from a cache drive to a cache pool.

Before migrating, I backed up my appdata using the plugin.

I created a cache pool with 3 cache drives including the original "cache drive". I switched from default (BTRS?) in cache drive to ZFS in cache pool.

After restarting array, I watched the appdata.zip backup disappear from the document share. (Maybe the appdata_backup.zip had never been moved from cache to array before moving cache drive into new cache array?)

I would like to recover and restore those backed up docker images.

I can live with recreating the docker images if the backup is gone.

I have no appdata share on my array after the cache mighration.

Are there instructions/guidance on restoring / recreating the appdata share?

Any suggestions for trying to find and restore the missing appdata_backup.zip file?

tower-diagnostics-20260925-1323.zip

Edited Friday at 08:38 PM2 days by OriginalFlavor

Your diagnostics show why Docker cannot start: both appdata and system still point to the old pool, cache. Your new pool is named cache_pool, so Unraid cannot create Docker’s storage directory.

What exact backup destination was configured in Appdata Backup, and did the backup finish successfully? From Main, browse the individual array disks and look for that backup directory. Do not limit the search to the guessed appdata_backup.zip filename. If you find a backup, copy it to separate storage first.

If the backup existed only on the old cache, it was affected by the formatting operation. Recovery cannot be guaranteed.

Once the backup is secured, or you decide to start fresh, create or edit the lowercase appdata and system shares. Set Primary storage to cache_pool and Secondary storage to None. Restore any recovered appdata while Docker remains disabled.

Then recreate the missing Docker image and reinstall containers through Apps → Previous Apps, if saved templates remain. Check any mappings that still use /mnt/cache/. Reinstalling containers downloads their software; it does not restore missing application settings or databases. https://docs.unraid.net/unraid-os/troubleshooting/common-issues/docker-troubleshooting/

Also, the current three-drive ZFS pool is a stripe without drive-failure protection. Consider the intended pool layout before restoring data.

1 hour ago, JorgeB said:

Also, the current three-drive ZFS pool is a stripe without drive-failure protection. Consider the intended pool layout before restoring data.

The purpose of the pool was to add more space and be faster than my array.

The cache pool is all NVMe SSD drives. The array are spinning HDDs.

Is stripe the fastest?

I think I am willing to risk redundancy for speed with the cace.

That's my reasoning. But I am listening.

1 hour ago, JorgeB said:

What exact backup destination was configured in Appdata Backup, and did the backup finish successfully? From Main, browse the individual array disks and look for that backup directory. Do not limit the search to the guessed appdata_backup.zip filename. If you find a backup, copy it to separate storage first.

If the backup existed only on the old cache, it was affected by the formatting operation. Recovery cannot be guaranteed.

It is not on any of the drives.

I did find an earlier backup from 2024. I had saved the current backup to the same directory.

Oh well. I think I figured out my weekend project--recreating Docker containers.

2 hours ago, JorgeB said:

create or edit the lowercase appdata and system shares. Set Primary storage to cache_pool and Secondary storage to None. Restore any recovered appdata while Docker remains disabled.

Then recreate the missing Docker image and reinstall containers through Apps → Previous Apps, if saved templates remain. Check any mappings that still use /mnt/cache/. Reinstalling containers downloads their software; it does not restore missing application settings or databases. https://docs.unraid.net/unraid-os/troubleshooting/common-issues/docker-troubleshooting/

I could not create the system share even though "system" was not on the list of existing shares.

However, after creating appdata, when I turned Docker off in settings and then back on, system share was created automatically.

Docker running great.

I did need to run "Docker Safe New Perms" to reach data with recreated Docker containers.

Thank you for your help.

Glad it's working again.

About the pool layout: yes, a stripe is the fastest ZFS layout and gives you the full capacity. But if any one of the three NVMe devices fails, you lose the whole pool, including appdata and the Docker image. For appdata and Docker, a single NVMe device is already much faster than those workloads need, so you usually won't notice the extra speed of a stripe

A raidz1 pool with the same three devices would still be very fast and would survive one device failure. Its usable capacity is about two times the smallest device. If you want to change the layout, now is the best time, because the pool is almost empty. Changing the layout means reformatting the pool.

Whatever layout you choose, set the Appdata Backup destination to a share on the array, not on the pool. Before you change a pool again, check that the latest backup is actually on an array disk.

Also, the 2024 backup is old, but for some containers restoring their folder from it may still be better than starting from scratch.