thoresson

On Unraid, zfs can be used in two ways: As proper zfs pools outside the array, and for single drives in the array. In the latter case, you don't get the self-healing features of zfs, but warnings if there are problems with the file system (a.k.a bit-rot). i just learnt that zfs can tell in what specific files problems surface. Is that true even when using zfs for individual drives in the array? If so, wouldn't that be a native way to get the functionality from the File Integrity plugin on the drives in the array: You run a scrub, get notified of files with errors, and restore them from backup?

I've never really understood how and when IOPS and R/W speed matters, and thought I didn't have to care given that my use cases are quite simple: I want to use my NAS as backup target for my family's computers and to run a couple of lightweight containers – but also as primary storage for 20 years of photos and videos, somewhere around 95k in total. And I just read a blogpost that hinted at that for my photos – and browsing folders with a couple of hundred photos over SMB – IOPS is quite important for how snappy the experience is. I am currently getting started, and have only got 2x1TB NVMe for a mirrored zfs cache and 2x4TB HDD for a one-parity array while I'm learning the basics. My idea was to build out the array with one or two more 4 TB drives, which would cover my needs for a foreseeable future. But if IOPS matters for my photo editing, would it be better for me to either expand my cache pool (getting one or two more NVMes and rebuild it as a RaidZ1) or to get two 2.5" SDDs and add a second mirrored zfs pool for photo storage?

Excellent! That allows for a great middle-ground for me: zfs for my most important data, where I also can make a better prediction for my storage needs, and the array as a backup target and some other use cases where guessing future need is harder!

If I configure a zfs mirror outside the array, will it get the same data integrity benefits from zfs as on any other zfs pool? There are some discussions on limitations on Unraid's zfs implementation around the forums and Reddit. But the way I read them, its about when you use zfs for individual drives in the Array? I'm considering a zfs mirror for my photo collection, which is the most valuable and irreplacable data I have, and storing the rest on Unraid's Array. Can Unraid spin down zfs pools like it spins down the Array? My "photo pool" will be used only when I do active work with my photos, so want that pool dormant most of the time. And finally, I if configure a zfs pool like that, I could also use that as a snapshot destination for my cache, which is also a zfs mirror. I wonder if that would mean I would dare to go with a single NVMe as my cache? My motherboard has 8 SATA ports, one m.2 slot, and one PCI x4. But because of shared lanes, when I use the PCI x4 for a NVMe adapter, like I do, I can only have 4 SATA drives. Should I go with Array + zfs pool, I would be limited to two drives in the array.

Thanks! Do you have any idea on how Rsync vs Rclone would differ for this use case?

Hi! Being new to Unraid and slowly learning, I try to understand how things works. And in particular when I have to change a default settings that could have security implications. To be able to access my Docker containers on the same IP addresses when connected to my home network as well as remotely using Tailscale, I've understood that I have to enable Host access to custom networks for Docker. But there is a warning to only do so if you understand what that means. And I'm not sure I do, yet. :) With it disabled, all services as well as Unraid OS itself doesn't seem to be able to communicate between each other unless the network is set to host. I guess this is part of the compartmentalization in Docker, isolating containers from each other. But that would mean that there also is at least some security aspect of enabling Host access to custom networks? Or are there other reasons this is off by default? If it matters, I'm not planning to expose anything on my Unraid server on internet, only remote access for me and my family using Tailscale.

Thanks! This world of expansion cards is more confusing than I had expected. :)

One use case for my new Unraid server is as the primary storage of 20+ years of digital photos. They are currently on a USB 3.0 drive formatted for MacOS, with a total volume of around 1.5 TB. What are my best options for getting them on my array and verify that all files have been correctly written? After some back and forth with Claude, I've this CLI command for Rclone rclone copy /Volumes/external_usb_drive /Volumes/share_on_unraid \ --progress \ --checksum \ --transfers 8 \ --checkers 16 \ --stats 5s \ --stats-one-line \ --exclude ".MYLock*" \ --exclude "*.tmp" \ --exclude "*.lock"Are there better alternatives, with either Rclone or other tools? And if I go with Rclone as above, should I do this subfolder by subfolder (photos stored in yyyy/mm folders on the usb drive) or am I good to go with everything at once?

I've just finished my first build and confirmed that everything works, now getting ready add the missing part to make it ready for 24/7 use: A second NVMe for a mirrored cache. Even though I got my first computer already in '94, this is my first endeavour into server territory and all the LLM bots fail to give me advice I can trust on this topic. So keeping fingers crossed for the wisdom of the forum! :) What I currently have is a SuperMicro A2SDi-4C-HLN4F with a 1 TB WD Red SN 700 NVMe as cache and 2 x 4 TB Ironwolfs as the array. The motherboard has only one NVMe slot, and one PCI-e x4 expansion slot. There are 8 SATA ports, but as they share lanes with the PCI-e expansion slot, as soon as I put a card there I'm down to 4 of the motherboard's SATA slots. I think I'l manage with 4 HDD's for a while, adding 2 x 8 TB and use one of the for parity would give me 16 TB of effective storage which I guess will be sufficient for my needs for a long time. But I also remember the huge storage capacity of my first computer... So, looking for a good future-proof way to add a second NVMe port that wouldn't limit my HDDs to max out at four drives. My initial thought was to try and find a PCI-e x4-card that have both some extra SATA ports as well as a NVMe slot but that doesn't seem to be an option. At least I can't find any. I had accepted that I would have to do with a PCI-e card that only give me an extra NVMe port and hope a 16 TB array would be enough for a long time. But browsing the forums for recommendations on such a card, I realised that another option would to get a m.2 expansion card for extra SATA port, like this one, and a PCI-e x4 card with room for two or more NVMe drives. But I can't find any cards like that either. What I do find are cards like this, with one NVMe slot and one for m.2 SATA. Could that m.2 slot be used with a m.2 expansion card for extra SATA ports? Any suggestions/recommendations appreciated!

Using Unraid as a backup target for the family's computers are one of the primary reasons I got a home server. We already do backups (three Macbooks, one Win11) to the cloud using Arq Backup. But I would like to add a local target to the mix, where the increased throughput would allow for more frequent backups. There are two conceptually different options: Using a backup tools like Arq Backup or Restic that does incremental, encrypted, and compressed backups. Using Rclone, Rsync, or something similar to keep a mountable copy of each computers home folder on the server. My initial thought was that the second option made most sense. On our home network, there is no need for encryption on the backup target, and having the files in a mountable share would make it easy for the other family members to restore individual files. But as I was doing research on wether Rclone or Rsync would be best for this, reading up on their settings, I realised that a "real" backup solution like Arq has a big plus: Built in versioning. It seems like that's possible to achieve with both Rclone and Rsync as well, with "archive folders" where deleted or modified files are stored. But with a complex folder structure and no easy way to configure rules for retention, the benefits I had in mind with "plain" folders are as big as I thought, and perhaps restoring older versions is actually easier using Arqs GUI compared to finding the right folder. So I'm know leaning towards Arq for local backups – but would like to hear what more seasoned Unraid users experiences/advice are. But going with Arq for local backups might also mean I have to rethink my approach to off-site backups: Here my plan was to take care of that from the server instead of from each computer, using Restic/Duplicati/Duplicacy/Kopia (interested in input for this decision as well!) to do off-site backups in a more centralised way. But if our computers' backups are stored as encrypted blobs, I guess these backup tools will have a hard time to work efficiently, with deduplication, versioning, etc. So, if I go with Computer -> Arq -> Unraid for local backups, should the computers manage their cloud backups individually and backups done from Unraid to the cloud only take care of data that is exclusive to the server?