Reuse zpool data

Hello,

I am using Openmediavault 7.

Following a ram problem I had major damage to my server. It was still working but with a lot of errors at startup.

But the zpool remained intact. It contains a lot of data.

So I exported the zpool then reinstalled the system and all the plugins, etc... Then I imported the zpool and the data is still there.

When I create shared folders, I reuse the folders already on the zpool. But it's as if there's nothing in them when I assign them to different services.

To be on the safe side, I've only done this with two plugins that can handle data.

I suspect that creating the shared folder will overwrite the data in the folder that was created on the previous system.

Or the data is not seen.

I would like to reuse the zpool data for services on the new system.

Can you help?

Thank you!

Normally the relative paths are correct, I had used the tree structure as you say (see screenshots).

I also tried to capture other information that might be useful.

What do you mean by datasets ?

My pool is divided into 4 Vdevs, a raidz1, a mirror of 4 disks, 2 log disks and a cache disk.

Thanks for your help.

Thank you for your quick and clear answers.

Unfortunately I created the folders directly in the pool.

Why do they have to create a datasheet beforehand?

It's a bit of a shame because the new openmediavault clearly sees the folders created previously and on the command line you can see that the files are in them.

And it worked very well like that before.

I can't create a datasheet because the data is already in the pool.

Unless the pool preserves and integrates the data, which I doubt very much?

Or create new folders in the pool and move the data manually into the new folder then delete the old folder but that's a bit barbaric and I don't know if it would work?

And there's 1.5 TB to move and the disks aren't ssd.

Is there any other way, or is it impossible ?

I understand the broad outlines, but it gets complicated at the end.

This part seems to concern the problem directly:

"This is because a ZFS pool will always have a mountable filesystem called the root dataset, which is mounted at pool creation. At creation, a pool is imported into the system, such that an entry in the zpool.cache file is created. At time of import or creation, the pool stores the system’s unique hostid and for the purposes of supporting multipath, import into other systems will fail unless forced."

My pool is mounted at the root without any particular mount point.

Next time I'll know that I need to create a mount point as you've described. And then use the new file system for the shared folders.

But you say it's good practice, so is it mandatory to transform data from an Openmediavault to a new Openmediavault?

My pool comes from a previous Openmediavault which crashed due to a ram problem. I exported this pool before reinstalling the system, plugins, etc, then importing it to the freshly installed system.

You say it's just a relative path problem.

When I create a shared folder, I give it a name, I select the pool (because I haven't created a file system from a pool mount point), I choose the path with the tree button, I see the folder already created from the Openmediavault that crashed and I select it. And I put a label on it.

Then I go to the corresponding service and select the shared folder I've just created which uses the correct folder in the pool containing the data used by the same service in the previous Openmediavault.

And then the problem arises - it's as if there's nothing in the shared folder.

The service has no data from the shared folder.

Yet the data was in the pool folder at least just before the shared folder was created.

Does the system overwrite the data in the pool folder when the shared folder is created? Does it not see them or something? If so, why?

Forgive me for my insistence.

Yes, the data is definitely in /POOL1

I don't have access to the computer but as soon as I get home I take photos of the folders in POOL1.

Thank you again for your

The first photo shows POOL1 and its contents. These are the folders created for the shared folders when they were running with the first openmediavault, which is no longer operational and which I had to reinstall.

The second photo shows one of the folders in /POOL1 (/POOL1/PHOTOPRISMIMAGES) and its subfolders.

The third photo shows the files in one of the sub-folders of /POOL1/PHOTOPRISMIMAGES.

When I assign these folders to the new shared folders for running the same services as before on the newly installed Openmediavault it's as if they were empty.

The services work as if the shared folders were empty.

As if there was no data inside.

I would like the Openmediavault services that I have reinstalled to use the data contained in these folders.

Because putting the data back into the zpool would take a very long time.

chente :

I had 8 containers on the previous Openmediavault.

After reallocating the corresponding folders from the old system to the new shared folder and after allocating the new shared folder to Docker Compose, nothing, no more containers (see screenshots).

All I care about is getting Photoprism to work - there are over 1TB of photos and videos.

I don't have any data on the other plugins, so I can't test with Samba.

After seeing what happened with Docker Compose, I'm very hesitant to do it for Photoprism for fear of losing more than 1 TB of photos.

I have snapshots of the old system and I can make new ones.

If I try with Photoprism could I go back?

Krisbee:

You're certainly right it's not optimal.

But that's not my question.

I have no intention of destroying this pool, which contains over a TB of data. And since we can't delete vdevs.

I could detach the log and cache and allocate 2 of the three disks to the mirror. Or create a new vdev and put them in.

Or just create a new pool?

But for what gain?

Knowing that the vdev, log and cache disks are not the same size.

As for the cached disk, I don't know if it's used.

I easily have 23GB/24GB of ram and a 70% processor. It's a 2 x xeon 6 cores (12 cores 24 threads) t 2.8 Ghz if I remember correctly.

My system may be oversized but it worked perfectly and for me that's what counts.

My aim is simply to reuse my zpool and its data for the services of the newly installed Openmediavault.

Thank you for all your explanations. It's a configuration that's not optimised because I have virtually no computer training. I've been struggling on my own for many years and can't count the sleepless nights. I make do with docs, forums (maybe not enough!) and tutorials.

The PC is a Z800, there's a 3.5" bay with 4 x 1TB 3.5" hard drives in raidz1. Another 2.5" bay with four mirror ssd, two 120GB (recovery) and two 240GB. Two 2.5" 120 GB ssd in log mode. And a 500GB m2 cache.

As for the performance problem between raidz and mirror, I suppose that the vdev raidz isn't slowed down too much by the vdev mirror since ssd's are much more powerful.

I recognise that the redundancy is not there.

I was aware of this, all it takes is for two raidz disks to fail or three mirror disks and the pool is dead.

Following your advice, the best thing would be to buy four 1TB 2.5" hard disks and replace the raidz and mirror vdevs with a single 8-disk raidz vdev, one of which would be a hot spare.

This would leave two slots for a docker pool.

As for the ZLOG, I didn't know that ssd's had to be power loss protected.

It wouldn't have cost much to add three poor 1 cent capacitors to all ssd's !!!

################################

Personal comment:

Fed up with the technology race, this Z800 is over 15 years old and yet it still runs! Isn't today's technology enough? and then we talk about ecology! Whether it's computers or cars, everything you buy will end up in the bin within 5 years...

################################

As for the cache disk, I can't see where to put it, it's an M2 ssd and it's mounted on a Pcie card. I don't think I can boot on it because the Z800 motherboard doesn't allow it, so it's not usable for the system.

Accessory question: I have a sas lsi card integrated into the motherboard and I've changed its ROM to ITmode passthrough. Does it do anything?

Normally at least 97% of the photos are already on an external hard drive, just the ones I recently added to the server. They're also on a commercial cloud.

I had used this external drive to inject the photos into the Photoprism import folder (/POOL1/PHOTOPRISMIMPORTFOLDER).

It took an extremely long time to load the photos into Photoprism. You have to transfer them to this folder while photoprism loads them and indexes the photos.

All this had to be done dozens of times because I'd repartitioned the photos into dozens of folders so that it wouldn't crash.

It was a real pain, which is why I don't want to do it again.

Hello,

Sorry for the late reply.

The problem has finally been solved. I must have misspoken or perhaps the Deepl translation isn't very good.

I had reallocated the folders corresponding to the Docker Compose data in the old installation to the shared folder in the new installation and then finally allocated them to Docker Compose in the new installation.

Result: nothing, no containers, no images, no volumes, etc...

That's why I was very worried about doing the same thing with the Photoprism shared folders.

I was very worried that the system would overwrite the data in the old shared folders when creating the Photoprism shared folders in the new installation.

I finally decided to do it late last night.

But chente was right, it doesn't overwrite the data in the old folders !!!

Photoprism works !

When the data disappeared from Docker Compose, I installed new containers that did not work, I just checked the access rights it seems good read / write / execute for root and users.

I'm going to delete all the old folders and the new shared folders and put everything back to new.

But it doesn't matter at the moment, I'm not doing much with these containers. The most important thing is that Photoprism works.

In any case thank you both !!!

I would have learnt a lot of things, datasheets with chente and optimising a zpool with Krisbee.

For the moment I'm staying as I am, it's taken me a long time to set all this up...

But I'll take note of your advice and implement it in the future.

Thanks again for your responsiveness, your efficiency and the time you've given me!

Have a great weekend!