samsausages

I had to deal with this again last evening after a reboot. May have a bigger issue here. If it happens again I'll report back, but usually don't do reboots often, so may be a while.

Great news! Glad it resolved it!

TLDR: The old docker macvlan network was bugged from the update. I had to remove it manually and recreate it. ---- Update on this, was looking at it more closely. I composed down all stacks & removed all docker networks to eliminate any potential conflicts and do more testing. On the docker settings page, I noticed that one of the networks, br2.10, doesn't have a gateway defined. Just shows -----. The others do have gateways assigned and when I attach containers it works. no way to edit that gateway, even with docker disabled. Resolution: Removed network manually with: docker network remove br2.10 (May give you an error, since it's bugged) Turned off docker. Checked Unraid Network settings and there still was an entry in the routing table for br2.10 Deleted entry manually. Went back to docker settings, now I could define a gateway IP for br2.10. Added entry. Enabled docker. Profit.

Going to 7.2.5 broke Nginx Proxy Manager for me as well. I'm using macvlan for the container. Tried ipvlan, same issue. Changed to dedicated docker network - works. With macvlan, I already had a MAC set in my docker compose file before updating. Tried without, or by changing it, no luck. No DNS resolution. ("communicate with host" is disabled) "getent hosts pypi.org" comes back blank. Docker container logs when on macvlan: [5/4/2026] [5:56:15 PM] [Certbot ] › ▶ start Installing cloudflare... [5/4/2026] [5:56:23 PM] [Certbot ] › ✖ error WARNING: Retrying (Retry(total=4, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError('<pip._vendor.urllib3.connection.HTTPSConnection object at 0x152a39cd0190>: Failed to establish a new connection: [Errno 101] Network is unreachable')': /simple/certbot-dns-cloudflare/ WARNING: Retrying (Retry(total=3, connect=None, read=None, redirect=None, status=None)) after connection broken by 'NewConnectionError('<pip._vendor.urllib3.connection.HTTPSConnection object at 0x152a39cd0e90>: Failed to establish a new connection: [Errno 101] Network is unreachable')': /simple/certbot-dns-cloudflare/

Is blackwell supposed to work? I just got a new card and it's not showing up. AI tells me blackwell isn't supported in unraid yet. Can someone confirm before I pull the card out?||| FYI, error is: NVIDIA-SMI has failed because it couldn't communicate with the NVIDIA driver. Make sure that the latest NVIDIA driver is installed and running. I don't see the card anywhere in the device manager. Have tried reseating it, made sure 4g encoding and r-bar are enabled. Just confirmed it works in another PC. 6000 Pro Blackwell GPU bertha-diagnostics-20260205-1437.zip

I'm used to configuring a firewall on all of my servers, as I like to restrict all inbound and outbound traffic/ports. But I don't see an obvious way to do this on unraid. I have been relying on pfsense for this, but not ideal as I don't want to have a vlan just to isolate the unraid server. Is there a plugin or firewall service that I'm overlooking?

I understand. Just trying to tell people what's going on, as the question has been asked here and it's how I found this thread and previous posts made it sound like stable does not work on Linux.

Just wanted to give an FYI on the DayZ Server. This docker container uses the experimental branch. Yes, Dayz does run on linux using the Stable Branch. However, the stable branch requires you to log in with your steam account. I confirmed this installing on a Debian VM. Instructions here: https://community.bistudio.com/wiki/DayZ:Hosting_a_Linux_Server#top-page

@limetech I do have a syslog server setup, but I don't have a good log aggregator setup, so digging through them is a bit of a pain right now. I did find the logs from last time I crashed, before I put the script in place. Attached are the relevant logs from right before and after that reboot. Looks like it crashed around 5am and I rebooted at about 5:35am If you need logs from another modules let me know, but pretty much everything else is just complaining about losing storage access. Nice to see a an attempt to fix it! I'm reluctant to try it right now because I can't easily roll back my zfs pools, so I'll probably just hang on unless that script stops working. unraid.log

@limetech No worries! With losing shares I mean that on the "shares" tab all the shares would be gone. Couldn't browse the shares even in the CLI and all containers/vm's that use storage crash. Doesn't matter if it's unraid array or cache. Disks would still show up on the "main" tab. I don't remember if I could still browse disks under /main/diskx, unfortunately. So I can't confirm if it was the fuse layer or the disk mount itself. A reboot would resolve the issue.

After losing shares about every 24 hours (did seem to depend on workload, usually when I ingested large amounts of data), I started searching for causes. There wasn't a specific error in the log so I had to go through trial and error. But my search lead me to people talking about the file limit, so I gave it a shot and I stopped dropping shares. My workload is mainly media, archiving and data hoarding, my stash is very large as I download entire youtube channels and entire websites, with hundreds of thousands of videos and well over 300TB on unraid. I have another server that runs Plex and connects to unraid using NFS. I also use a container that generates plex thumbnails in parallel, so it's faster, and a bunch of other services that process media. Running all those jobs often resulted in the array getting dropped. It's a 3rd gen Eypc server with 40tb in NVMe, and I do tax the system quite a bit when I have other systems accessing it to help process media. I did see it, rarely, without NFS. But it was so rare that I didn't worry about it, maybe once in 6 months. But when I moved my Plex to a 2nd server and added NFS for the file share, it got a lot more frequent, and when I started using containers to help re-build the plex thumbnails, it became a daily occourance.

I recently ran into a similar issue when adding an NFS share between unraid and my proxmox server. I'd lose shares after about a day. Looks like my issue was caused by unraid hitting the maximum open file limit, only set to 40k by default. I'm using this script to increase the file limit, haven't had any issues since: https://github.com/samssausages/unraid_scripts_and_fixes/blob/630629b0c40309f93626518f9f78472aa36fcc2c/unraid_increase_file_lock_limit.sh

The current version has been working great for me. If there is something you're missing let me know.

Kicking around the idea of replacing my parity drives with some dual actuator HDD's. Wondering if anyone has tried it yet and how performance is looking for them.

Feel like LT catching flack for Broadcom's BS. People really are on edge out there!

Thanks for the update. Pretty much what I expected from LT, support for existing community. You all haven't let me down yet.

Unraid hasn't let me down yet. I'm not going to worry or get upset about something that hasn't even been officially announced yet, comes from Reddit speculation on a Sunday night into a Federal Holiday... Unraid and LT have build more credibility with me than that, over many years. I have a feeling license changes are coming with the talk of increasing the array disk limit to exceed 30. That would explain the extra tiers. i.e. unleashed. But as of right now, I have no reason to believe that my multiple Pro licenses won't continue to be honored as advertised. I do understand the need for model changes, as I paid for this once and I have been getting updates for many years. Now I do expect them to honor our agreement (lifetime), but everyone should be able to make new and unique license terms as they see fit, going forward.

I'm also looking for nvme-cli as I have some Intel NVMe SSD's that I need to low level format from 512 to 4k. Would be nice to have it in unraid by default, or in nerdtools! But I did find the package online and installed it manually, sure was easy enough... took me longer to find the package than it did to install it.

I manage file permissions at the linux level, that is my intention for my use case. I don't always want the SMB user to be the owner of the file. I also don't run all my containers with 99:100, sometimes I use masking, sometimes I don't, depends on what I'm doing.. I'm not telling other people to do what I'm doing... I'm asking what others are doing so I can get ideas for my use case.

So for a while a lot of my SMB files were changing owner to the SMB user I was using to log into SMB Unraid through Windows. I added some lines to the SMB config file that forces the default user to "nobody:nogroup" I'm using this and it works fine defaulting it to nobody:nogroup [share] path = /mnt/user/share public = no browseable = no guest ok = no writeable = yes read only = no force user = nobody force group = nogroup create mask = 0640 directory mask = 0750 #Show ZFS Snapshots vfs objects = shadow_copy2 shadow: snapdir = .zfs/snapshot shadow: sort = desc shadow: format = autosnap_%Y-%m-%d_%H:%M:%S_daily shadow: localtime = yes However, I think it would probably be better if I had it inherit the user from the root folder instead, as I have some more restricted folders setup. But this doesn't work as expected. Removing "force user" "force group" and then adding this has no effect: inherit permissions = yes Is there a better way of accomplishing this?

@NeoDude I don't think this is a zfsmaster issue, but I think I got an idea of what's happening. What is your folder and dataset structure? You may need to execute the command with -r "recursive" It sounds like you are trying to rollback the parent dataset and it has children datasets nested within it. When you browse .zfs/snapshot of the parent, then you won't see data for the children. But if you go to .zfs/snapshot in the children, then you will see the data of the children. The only way for you to mess this up is if you didn't take your snapshot with -r recursive. If you have the snapshot, then the data is in there. Another reason I can see for this is if the dataset isn't being mounted. To restore, you need to do the restore on the specific child dataset. If you have subfolders within, that are actually configured to be datasets, then you need to restore each folder/dataset. (Or make sure it's doing it recursively) To pretty much clone the dataset, recursively with all the snapshots and properties you use: `zfs send -wR metapool/appdata@migration | zfs receive -Fdu metapool/appdata_new` But read my notes below so you understand the flags, as that's unmounted. and name the send differently for your testing, so you don't override anything. Then when you confirm the data is there, you can rename it and update your mount points if needed. Here are some of my notes on how I have done it in the past. Hope it helps you! If you need more help you can IM me so we aren't clogging up the thread here. # Backups & Snapshots ## Snapshots ### Create New Snapshot `zfs snapshot workpool/nextcloud@new_empty` Recursive: `zfs snapshot -r workpool/nextcloud@new_empty` ## Transfer Dataset from one location to another ### Create Snapshot 1. `zfs snapshot -r metapool/appdata@migration` ### Send to new dataset (Recursive with dataset properties & snapshots) 2. `zfs send -wR metapool/appdata@migration | zfs receive -Fdu metapool/appdata_new` -w sends raw data, needed with encrypted datasets. also keeps recordsize & options. -R is recursive and includes all snapshots/clones -F Forces the overwrite of the target dataset - use with care! -d Uses the provided dataset name as the prefix for the names of all received datasets. Essentially, this means that the data will be received into the named dataset, but not as a clone -u ensures that the received datasets are not mounted, even if their mountpoint properties would typically cause them to be automatically mounted. Requires manual mounting! ### 3. Confirm data is in location and present 4. Rename old dataset to appdata_old 5. Confirm mount points changed for appdata_old 6. Rename appdata_new to appdata ### Mount Dataset (if left unmounted with -u flag) 7. `zfs mount metapool/appdata` Done ## Syncoid ### Replicate Dataset from dataset (Not fully tested) `syncoid metapool/gitea/database workpool/gitea/database`

@Iker I can't tell you how nice it has been to use the plugin since you changed the refresh methodology! Huge improvement for my use case! I do have a future feature request: The ability to refresh by pool. I.e. a refresh button on the pool bar that has the "hide dataset" "create dataset" buttons. And/or in the config the ability to select/deselect pools from the refresh. How I use this: All my ZFS pools are SSD's, so I don't care about spinup/down on those. But I do have some ZFZ formatted disks as snapshot backup targets in the Unraid Array. I rarely browse those and don't need to update ZFS Master very often. Being able to exclude just those pools (or having a button to refresh only the pool that I'm working on) would make it so those zfs array disks spin up even less. But even without that, it has been a huge improvement! My disks are only spun up for about an hour a day now, where before they spun almost all day!

@JorgeB I guess I better test it again and monitor my ARC more closely, I didn't check my hit rate. The read test I ran last night was giving me 17,000 MB/s-20,000MB/s, on my VM with a ZVOL and primarycache=none So I figured that performance is coming from the ARC.

Updated to v0.9. Lots of upgrades to make it more durable and robust. Mainly error checks and validation checks.