As a noob, I am a bit concerned about migrating to Scale, can someone calm my nerves?

[asap2go]

I went back to edit my post to remove the part where I said k3s will use 10% (or X%) of your CPU while idle, but couldn’t find where I said that.

Low-end or low power systems, of which there are NAS users, need to be aware of this for SCALE.

I've seen power usage drop from 190 to 180W after migrating to scale (complete rack including switch with PoE, etc.). Probably due to better hardware support.
In the end no one cares about unused features but overall performance and efficiency. Well now a bunch oft my RAM is just idling around but that's another story :D

 

[William Bravin]

That depends on which functionalities you mean:

• ZFS improvements? Yes, that's the plan.
• UI improvements? I hope so, but I don't recall hearing either way from iX
• Clustering? AFAIK, there's no plan or intention to do this.
• Apps? Same.
• VMs? Not really possible for the most part.

...which is why many of us suspect that iX plans to drop CORE entirely. As yet they deny this.

Thank you for responding

In the event IX decide to kill core i would move to scale.

If it will get to complicated for me I and i cannot use core (I truly hope not) if so I will most probably migrate to unraid

 

[sfatula]

I went back to edit my post to remove the part where I said k3s will use 10% (or X%) of your CPU while idle, but couldn’t find where I said that.

Low-end or low power systems, of which there are NAS users, need to be aware of this for SCALE.

Agreed. And I don't think you explicitly said that in this thread, but you did link to some. Anyway, we do have a lot of people who come here with ancient systems, laptops, etc. and want to run Scale for some reason.

I'm actually using a Raspberry Pi as my desktop, for power reasons. Works great and is plenty fast, spreadsheets, whatever. But no, I don't run Handbrake on it, I do that on Scale with the container. That part would be slow. Since I follow the Pi forums and reddit, I know they try and run insane things on their Pis. So, it can be an issue.

I wish (maybe there is but I don't use it as Emby is much nicer ) there was a Plex OS so all these Plex users can have an easy to install and use system, for me, if they know nothing about Linux and they want to install Plex primarily only, I don't get it, why Scale? And maybe their *arr apps as an addon in the custom OS, kind of like what Home Assistant did with their OS, or, Kodi with their various incarnations that actually run well on a Pi.

 

[winnielinnie]

The thing is, "Apps" on SCALE, of which the backend is k3s, can be an analogy for anything in the "low-end systems" paradigm.

Take for instance "Transcoding". Erase everything about k3s. Forgot technical details and specifics. Forget metrics.

There's a thing called "Transcoding", and it's a "feature" of some software. Let's say Plex, or Jellyfin, or Emby, etc.

It's a "thing". It's sexy. Some people love it.

It requires at minimum some sort of decent hardware, especially for higher resolution videos.

My advice, casually speaking human to human would be this: If you rigged up a low-power system to stream media, you should consider converting your media (ahead of time) to a more friendly format for your household devices, or even test out direct streaming without conversion. Don't just migrate to this thing called "Transcoding". It has minimum CPU/GPU demands, and you might have a poor experience with it.

Is it true that transcoding video on-the-fly is not the same for every system? That high-end systems with impressive GPUs can handle it easily, without affecting other users working on the same system? Sure.

Is it true that its use of system resources can vary, depending on many hardware and software factors? Sure.

But the advice remains the same. Avoid relying on transcoding if you have a low-power system. There are other alternatives.

 

[William Bravin]

Agreed. And I don't think you explicitly said that in this thread, but you did link to some. Anyway, we do have a lot of people who come here with ancient systems, laptops, etc. and want to run Scale for some reason.

I'm actually using a Raspberry Pi as my desktop, for power reasons. Works great and is plenty fast, spreadsheets, whatever. But no, I don't run Handbrake on it, I do that on Scale with the container. That part would be slow. Since I follow the Pi forums and reddit, I know they try and run insane things on their Pis. So, it can be an issue.

I wish (maybe there is but I don't use it as Emby is much nicer ) there was a Plex OS so all these Plex users can have an easy to install and use system, for me, if they know nothing about Linux and they want to install Plex primarily only, I don't get it, why Scale? And maybe their *arr apps as an addon in the custom OS, kind of like what Home Assistant did with their OS, or, Kodi with their various incarnations that actually run well on a Pi.

Hi all

I was using emby on core and when i installed scale all hell broke loose. I since migrated to jellyfin (a breakout from EMBY). I never liked Plex.

On my Dell r720 running proxmox 8 I have Truenas core (not scale) Home Assistant OS and OPNsense as VMs.

I will be adding a windows vm to run kodi on this server (yo allow remote connections)

Since both OPNsense and HA are always running I decided to install them on the Dell.

I figured that the dell albeit uses more power that the PI it would still use less than if i were to have seperate PI or pcs for each solution

 

[Whattteva]

On the other hand, merely posting that because you read some people having a specific problem does not mean that the problem is what you say it is either. It is indeed an issue or problem perhaps. In this very thread, someone who migrated to Scale from Core had a higher CPU usage, which you then immediately assumed was k3s (could have been, or may not have been). Maybe the migration has a bug, maybe if he did a new install it wouldn't. In that case, the actual prolblem is the migration right? I can cherry pick problems people are having on freeBSD and state that freeBSD has some issue too as a few users had some similar sounding problem. That wouldn't make it valid. I can find thousands of issues posted here and make broad claims. That is what I feel you are doing here. I have heard few people complain of that, the last time someone did they found out that it was one of their apps, just last week, the thread was titled something about high CPU usage. Problem was solved. High cpu usage can be all sort of things, you want to associate that with Scale and k3s. But yes, indeed, having one more layer is going to add something to resource usage, we all know that, and that will be worse on lower powered machines. Undoubtedly. But Windows has more resource usage than dos too. As things progress, more power is needed but it doesn't make it a problem per se.

Ok, that's valid. But that IS a fairly frequent complaint seeing that most people that post about it here usually run SCALE for apps and there are numerous ones complaining about the dreaded apps stuck at "Deploying" forever. I do concede that yes, I may have painted with too broad a brush, but it is still a fairly prevalent problem, nonetheless.

The linked to blog post is light on details. The github post is about small (embedded) systems by the description, and many of the linked to cases are as well. Not very compelling except if your point is k3s will take more resources than no k3s, well, yes, that's true of any software. Stating things like high load average is conflating what load average is vs cpu percent, etc. Note the issue is not reproduceable indicating it's something on the specific machine too. The github issue is full of Rapsberry Pi users (I use one) and that is not surprising at all. Yes, k3s undoubtedly uses > 0 resources and may be too much for a Pi. Note the very last post on github, the guy discovers it was a specific helm chart. It's hard to say what the actual problem is for all the other folks. The most compelling of your links was actually to a truenas user with lots of k3s processes. That was clearly a bug of some sort with Scale.

I posted one link, but as I said earlier, there are many threads about it, quite many of them even with Xeon's, which obviously are leagues beyond Raspberry Pi in performance. This one here has a guy saying that he has seeing 20-25% on one of his Xeon cores constantly.

In any case, what this is about is someone is having some sort of problem with cpu usage. What that is is unknown except for the solved cases on github that claimed this and turned out to be something else. It could be k3s, but then you would expect far more posts about that in my view here on the truenas forums which I've seen very few of (yes, more than 0). And on github, people will chime in and say they have the "same problem" when it is totally different, this is not surprising or unusual. High CPU by itself means nothing as far as determining why.

A trend of significant number of users experiencing similar issues can be FUD, but is rarely the case... at least in my experience developing software. We developers like to avoid addressing it because it's often hard to reproduce, but it is almost always an actual issue that needs to be addressed.

I realize there is a lot of hatred (or bias against) here for/against Scale/Linux and I don't want to discuss that but many posts in these forums have that. Core is a great product and undoubtedly it's still more reliable, I believe we can agree on that. That will change over time. When you say things like "this is why I will never install Scale", Scale is Beta, etc. that kind of indicates where comments are coming from. When I read things like that, my first natural impulse is to think I will reduce the value of the entire posters comment due to bias. That's the natural inclination, but then I realize I shouldn't really.

This is fair. I do definitely have bias against Linux. Thanks for pointing this out and I will try to avoid inserting my own biases in my posts.

Really to me, Scale is more complex to use than Core. I don't mind that as a retired It guy as tech is trivial for me, and thus my Scale machine runs perfectly with no issues really. But so many home users who have limited docker / Linux knowledge is definitely going to create a lot of posts with various issues. That will take a while yet to prevent.

I'm glad you are here volunteering your time, you have a lot of good posts and knowledge to bring to the community in general. I just see this differently but appreciate your view. I just wish there was less of Core vs Scale in these forums. And that's what I saw your comments as. I could well be wrong about that. But pretty sure I'll get a lot of blowback on this post!

I definitely disagree with your stance that k3s isn't the issue that is so often experienced by users on TrueNAS or other platforms using it, but you definitely also brought up a legitimate issue with my biases and painting with too broad of a brush. No blowback from me here.

 

[asap2go]

The thing is, "Apps" on SCALE, of which the backend is k3s, can be an analogy for anything in the "low-end systems" paradigm.

Take for instance "Transcoding". Erase everything about k3s. Forgot technical details and specifics. Forget metrics.

There's a thing called "Transcoding", and it's a "feature" of some software. Let's say Plex, or Jellyfin, or Emby, etc.

It's a "thing". It's sexy. Some people love it.

It requires at minimum some sort of decent hardware, especially for higher resolution videos.

My advice, casually speaking human to human would be this: If you rigged up a low-power system to stream media, you should consider converting your media (ahead of time) to a more friendly format for your household devices, or even test out direct streaming without conversion. Don't just migrate to this thing called "Transcoding". It has minimum CPU/GPU demands, and you might have a poor experience with it.

Is it true that transcoding video on-the-fly is not the same for every system? That high-end systems with impressive GPUs can handle it easily, without affecting other users working on the same system? Sure.

Is it true that its use of system resources can vary, depending on many hardware and software factors? Sure.

But the advice remains the same. Avoid relying on transcoding if you have a low-power system. There are other alternatives.

I can't really relate that example to why BSD or Linux are better in a low end system.
Linux has better hardware support.
Better support for power management, and so on.
Just claiming BSD is more efficient because it uses less ressources out of the box is just wrong.
With that kind of argument one could claim ARC is better under Scale than Core. Uses less memory after all.

If we talk stability and stuff then sure.
FreeBSD uses all of my RAM and I never had an update of Core cause major issues for me unlike scale. But that's not really a Linux/BSD issue.
And it will be a temporary problem considering the allocation of development resources.

This is really the only forum where Bhyve vs. Kubernetes is even a discussion.
VHS has won. For better or worse ;)

 

[Whattteva]

This is really the only forum where Bhyve vs. Kubernetes is even a discussion.
VHS has won. For better or worse ;)

You're wrong. The cloud streaming services have won . Any kind of media of any form has lost, including USB sticks.

 

[sfatula]

But the advice remains the same. Avoid relying on transcoding if you have a low-power system. There are other alternatives.

Absolutely true, though I don't ever want to transcode anything realtime. Not to mention that software encoding like with ffmpeg is superior to hardware encoding anyway, it's just slower. But the same people who want that awesome quality, end up transcoding with a GPU!

All my media is in H.265. Direct played. My clients are Apple TV, Roku, and iPad, all of them direct stream.

 

[sfatula]

Any kind of media of any form has lost, including USB sticks.

They still make those?

The apps deploying thing is usually either the old setting in the config with hostpath, OR, a misconfiguration and they don't know how to check the logs to see what the real problem is. Correctly configured apps start.

We actually all have biases, very difficult to keep out of all posts. I don't like "apps", too limited mostly, I like running the containers themselves in Scale (though under kubernetes but less to no limitations). One of my biases.

Update: added one more. Or, apps fail to start after an upgrade from Core to Scale it appears, or even to Cobia sometimes.

In the past, on any OS or platform or appliance I used, I would never ever do updates. I'd fresh install for this very reason and still do. Even on PCs. Yes it's more work but on most any platform I've ever worked on, avoids issues. The extra time though gives me a clean machine, even my junk is cleaned up. Just takes a little extra time. In Truenas, this is pretty easy given the whole OS is imaged in. Configuring the machine doesn't take long, and, setting up apps is maybe 30 minutes as I have the settings all documented. Done this on Widows, Mac, freeBSD, MPE/IX (oops you guys don't know that one), Ubuntu, Debian, SCO/Unix (another oldie), Redhat all come to mind.

 

[Primož]

I'm slightly rethinking the decision to migrate to SCALE given some issues I'm experiencing. I've been running a CORE system for almost six years now but was forced to do some work on it as one of the drives in my array died. I made a post about migrating on reddit:

I am looking for the best way to transfer data between two separate machines.

About three months ago one of the three HDDs in my Z1 array in my NAS failed, forcing me to finally bite the bullet and splurge on an upgrade of drives (3x6 TB Z1 to 2x20 TB mirror). Luckily I have a box that was unused and I could put the two drives in there and spin up TrueNAS SCALE - the intention was to move from Core to Scale for the app support (Linux and all).

I managed to replicate the whole array using a replication task (pull direction and all), but I'm having problems with permissions on the new array - datasets seem to be read only, they are not case insensitive (SMB share ready), so I can't access them via an SMB share, when I can access new datasets and shares and I can access them same files on the old array.

I would also like to change the structure of my datasets a bit (there are a few nested datasets that I would prefer to unnest).

Syncing snapshots doesn't seem to be the ticket as it gives problems with permissions.
Rsync?
I was thinking about dropping the drives into the original box and CLI copying over the files with all the attributes (file creation and modification dates and the like) and start fresh snapshot wise, but I suspect setting up all the apps will be quite a chore and I would prefer run the existing server for the time being (with a degraded array with data backed up) and have the new data layout to set up all the apps?

Would plomping the drives in the old server, syncing stuff, taking them back to the new box, setting up Scale, taking them back and rsyncing them do the job? Any guides on how to copy the data via the network through CLI and keeping the attributes?

Endgame is to run Scale on the old box with transmission, plex and a few other apps with all the old data as is.

Any other suggestions?

The thing is... Core basically works. The 'only' reason I want to try out Scale is the Linux part as Plex will support sonic analysis & co for music (we have a long topic on Plex forums and BSD support for it) and the support for the likes of overseer, lidarr/radarr/sonarr & co to automate my plex server. If Core supported a bare metal hypervisor, running a Plex server in a virtualized Linux OS in a docker container would be easy. But having tried it it's a no-go, the performance is ABYSMAL, even for music only content, let alone for video. That's why Scale is so attractive... But Core mostly just works...

 

[sfatula]

The thing is... Core basically works. The 'only' reason I want to try out Scale is the Linux part as Plex will support sonic analysis & co for music (we have a long topic on Plex forums and BSD support for it) and the support for the likes of overseer, lidarr/radarr/sonarr & co to automate my plex server. If Core supported a bare metal hypervisor, running a Plex server in a virtualized Linux OS in a docker container would be easy. But having tried it it's a no-go, the performance is ABYSMAL, even for music only content, let alone for video. That's why Scale is so attractive... But Core mostly just works...

I would maintain, as a year+ long Scale user, that Scale works also IF you follow some simple practices. I have zero issues on Scale. But I don't do many things others do. I don't update to .0 or even .1 releases (Cobia), I reinstall instead of upgrade, etc. I try and keep things as clean as possible.

 

[Whattteva]

I would maintain, as a year+ long Scale user, that Scale works also IF you follow some simple practices. I have zero issues on Scale. But I don't do many things others do. I don't update to .0 or even .1 releases (Cobia), I reinstall instead of upgrade, etc. I try and keep things as clean as possible.

I guess that's a good strategy especially considering that's kind of the point of TrueNAS being "appliance/firmware"-based with the config file.

 

[Upuv]

My Two Cents,

I'm a long time NAS user. I've had many NAS''s over time. Here are the major jumps. There are other systems.
1. 2006 DIY linux host running SAMBA for the home network.
2. 2009 QNAP 6 disks. Rock solid box until it wasn't No path to recovery.
3. 2016 Synology Upgraded to this in time as I was worried about the QNAP. I managed the migration just in time.
4. 2023 TrueNas Scale Bluefin. AMD Ryzen 7 5700G

Now I jumped to Scale for a lot of reasons. Note: A lot of this would apply to Core as well.
1. The price performance difference between build your own and the big vendors was large.
2. VM's, Containers on host.
3. Linux rather than BSD. Hardware support was the main driver. Comfort was the other.
4. ZFS. Finally something that is actually portable. I can pull the disks and plop them on another system and I can just load up the pool. Happy days.
5. Did I say VM's. Yeah this was a big deal.

Now the experience had some learning to it. Getting the bios flashed on my HBA card was a royal pain. Lots of incomplete and conflicting doco on the net for this stuff. This is not a Scale problem. Bashing my head against the wall when setting up VM's and thinking networking was completely borked. Only to discover that it's basically mandatory as a home user to set up a network bridge. ( Insert flame from people about this is not a recommend Enterprise config. ) Once you know this it easy. Finding out that bluetooth and wireless are not supported was annoying. Would have been nice to use bluetooth to get the the console from my multipoint keyboard. And not having a wireless nic on the NAS did close a few options on my home lan. But easily compensated for.

Now most of my gripes are because this is basically Enterprise gear used in the home. Overall having a much more powerful compute on the NAS has done a great job of consolidating compute in my home. Those persistent small thing that I had running on NUC's, VMS, Pi's etc scattered all over the place have now contracted onto the NAS. My older NAS's could never really support more advanced compute needs. This one will last for years. And the storage can be easily migrated to the next motherboard and CPU at will. It's done a wonder for cleaning up the wire nightmare hell.
I still have other storage on the network. But they are all relegated to essential replication tasks. And these are about to be made redundant by doing a cross site replication with the brother in law. ( Once we can agree on the VPN set up. )

TL;DR
Yeah it was worth while for me to make the jump to TrueNas Scale. ZFS is the biggest thing as it provides the safety.

 

[Etorix]

Now most of my gripes are because this is basically Enterprise gear used in the home.

This is another key point: ZFS is enterprise software. You can't get the safety without the investment in decent-grade hardware and in learning enough of the basics to be your own home ZFS adminstrator.

 

[PhilD13]

I'm also a long time NAS user.
1. Qnap 8 disk with 12 disk expansion added. rock solid and worked perfect until it didn't. System old enough they quit updating the QTS software
2. Qnap 12 disks 2 years younger than previous box. Rock solid until December 28 2023 when it killed an entire row of the backplane, and crashed 3 of the 6 drives. Drives from that row may be damaged, they are in watch very close mode and to be replaced as funds allow.
4. July 2023 Truenas Bluefin on Supermicro chassis and backplane - Okay but little things pop up software wise that worry me about reliability
5. Dec. 28 2023 Truenas Cobia - Forced to make a new Truenas box due to the second Qnap suddenly dying.


I went to Scale for a few reasons. And yes I complain sometimes about broken stuff in Scale, but I still use it.
1. The price difference between build my own nas and the big vendors is huge. $3500 without drives vs $1200 without drives for more drive bays.
2. Linux. Hardware support going forward should be better.
3. ZFS. Like the filesystem overall and it should keep getting better.
4. Ix systems seems to want to go in the Scale direction with Debian underpinning things. Good enough for me.
5 I trust it just enough to make it the main focus of my storage systems on the premise it will become better as time goes on.
6 Core used in enterprise. Scale eventually will be too.

I'm not sure I would recommend at this time to migrate from Core to Scale unless you are willing to manage the fires after each update. It's just not mature enough to set and forget and is a work in progress. Maybe in a year or so yes Scale will be better so if core is working for you fine, then keep using it.

 

[William Bravin]

Hello all

Just as an update, i did migrate to scale as my main nas and relegated my old nas (supermicro) running core, to a backup server.

All seems to work fine so far including Jellyfin server jail on scale.

In addition i'm proud to have achieved backing up OPNsense Home assistant and proxmox to the truenas media

I feel there still several tweeks that will need to be made to scale to be 100%
Namely
Modify the gui to fit one's needs (not urgent)
alignment of menus and submenus (nice to have)