What is Unraid and how to build an Unraid media server

[Manni]

Hi Paul, Hi Jamie,

Awesome thread! I read it from the beginning and learnt a lot, installing quite a few plugins as I went along... The discussion on the shares, split levels and issues when copying is going to save me a lot of time.

As discussed recently with Paul, my Qnap NAS has gone bad just because a $10 ethernet interface has failed, with no way to repair it.

This has decided me to move to unraid. Luckily I have another QNAP NAS that hasn't died yet, so I'm going to copy its data to the unraid (so that I have a backup), then move the discs from the dead QNAP to the living QNAP. According to Qnap support, I should be able to read the array fine.

For now, I've repurposed my old HTPC as an Unraid server (I took the 1080ti out and replaced it with an old 7600GS). This old HTPC had a decent Antec case and already a 5-in-3 bay. It's a 3770K, 16GB RAM, Asus P8P-Pro motherboard. I installed an Adaptec SATA card with 16 ports, and I managed to fit 12 HDDs (old 4TBs I had in my Synology NAS). The case has lots of ventilation and the drives run cool, so it's not crowded.

For now I've only received two 4-port cables for the Adaptec, but I've managed to use 4 SATA 6GB ports from the MB for the remaining 4 drives. I have 10 data drives and 2 parity drives. I'll add my 1TB Samsung SSD as a cache once I've moved all the data. I love the fact that I was able to build a 48TB NAS with components I had lying around. The only expense was the Adaptec SATA board and the 4 sets of cables... Very satisfying!

Besides the media duties, I plan to run a VM with MyMovies and a single CCC instance, as discussed with Paul, as the UNraid server will always be on if I'm doing anything MyMovies related. I also plan to use it as a bridge so that I can always control my Oppo 203 clone with the MyMovies iOS App.

I only have two questions for Paul at this stage (if I may):

1) What is the name of the Server Layout plugin? I couldn't find it and it would be a lot better than the Excel sheet I use to track my HDD locations, warranty etc.
2) Once you've created a Windows 10 VM, how do you actually install Windows / access the VM? I don't mind using Remote Desktop, but how do you know the settings to access the VM before you've even installed Windows?

A small contribution to the thread, as it's a tip I use when I transfer large amounts of data (NAS to NAS or when backing up a large array to hard drives):

1) I always create a @From subfolder in the source share on the source array and a @To folder in the destination share in the destination array.
2) I move the first chunk of data (subset of x titles) that I want to copy in one go to the @From folder in the source, so that it's separate from the titles that have not been copied yet. If using Windows Explorer, I make sure the folders are sorted by name BEFORE the copy, I select all the titles in the @From folder, and I "drag and drop" them into the "To" folder, clicking on the folder at the top, to make sure the transfer is done in an alphabetical order, starting with the first folder, one folder at a time. If you don't select the top folder when drag&dropping, Windows will start the copy with the folder you have selected, which doesn't help if something goes wrong.
3) I copy these titles to the @To folder in the destination array. That way, if anything goes wrong, I can immediately see in which folder is incomplete (the last folder in the @To folder in the destination in the destination share), delete that partially copied folder and restart the copy from that folder until the whole batch is done.
4) Once that batch is copied properly, I move all the folders from the @From folder in the source share to a @Done folder in the source share, so that I now have a different folder that only contains the titles I know have been moved properly to the destination share.
5) I move all the correctly copied titles from the @To folder in the destination share to the share itself, so that the @To folder is now empty again.
6) I select a new batch of titles from the main share and go back to 2) to copy a new subset of titles.
7) Rinse and repeat until the whole share is moved / backup up (source share empty, all the titles in @Done on the source, all the titles in the destination share, @To folder empty).
8) If I plan to keep using the source array, I move all the titles in the @Done folder back to the source share.

I plan to do this if I decide to use a Windows client to move the date, otherwise I'll use some of the methodology if I decide to use a direct connection (assuming the performance penalty is too big with the Windows client).

I'm also setting up the Unraid server so that it goes to sleep with S3, and I plan to use WOL to wake it up when necessary. I'll let you know if that works or not, but it should. I've set the HDDs to go to sleep after 15 seconds.

I've set up a test array, it took a bit less than 9 hours to build the 2 parity drives, I'm going to run some speed tests (test performance with/without parity) and various reliability tests before I fill up the array. I didn't run a pre-clean on the drives because they were "pre-prepped" using WD Diag (write zero on all discs from my Windows PC, which I did when I de-commissioned them). They all passed that test, so I didn't feel like running a pre-clean as well, but maybe I should zap the array, pre-clean all the disks and rebuild from scratch.

Anyway, if this experiment / emergency rescue mission works, I plan to get a 24-bay case (they are out of stock at the moment) and create another UnRaid server with an almost identical motherboard I already have that uses an i5 2500K and also 16GB of RAM. Or maybe I'll swap the motherboards to get the better system on the 24-bay. I'll use the 6TB drives I have in both QNAPs on this, so 16 x 8TB in total, I should be good for a while, as I also have 24x6TB in my Synology DS2411 (with DS1211 extension)...

Ultimately I want to get rid of the non-Unraid servers (or keep them as backups so that I don't lose anything if they die) but it will take a while as I don't really want to buy 24x12TB drives right now...

[Manni]

Just to say that I found the Server Layout plugin. I had to use Google as for some reason it's not in the repository app.
S3 sleep works in manual mode, WOL works perfectly, but unfortunately S3 Mode after inactivity doesn't seem to work.
Anyway, my only question is about VMs but I'll Google that, I'm sure I'll find the info I need.

Happy holiday everyone!

[Pauven]

Hey Manni, welcome to the light side. After over a decade of using Unraid, I could never imagine using anything else. There's definitely a learning curve, but it's worth the effort for such a unique and awesome product.

From your write-up, I assume you're repurposing drives from your existing NAS units, right? You mentioned going for a 24-bay enclosure in the future, which is why I assume this is your path. I don't know how much capacity you need, but there is some benefit to starting fresh with the current large HDD's capacities available.

For example, my 24-bay case is chock full of drives, no slots unused (2 parity, 19 data, 1 outside the array, and 2 slots available for pre-clearing operations). Yet I only have 87 TB of total storage (never thought I'd be able to utter that sentence...). On the flip side, if I was building a new server today, I could pick up 6 of those new 20 TB drives, and that would give me 100 TB of storage. I could have a much smaller case for bigger storage. The need for dual parity goes down with fewer drives, but if I wanted even more protection then I could add a 7th drive. With my current 19 data drives, I consider dual-parity a must-have, but with only 5 drives I think you could get by with single parity.

Your collection is bigger than mine, so I assume 100TB is not enough, but the point is the same: larger capacities are actually more cost effective, as you need less case, less bays, less controllers, less cables, less cooling, less power. Less is more, in this case. There's also the "upgrade penalty effect", which is what I went through a few years ago. I built my server with all 3 TB drives (most cost-effective option at the time), and when I ran out of space, I jumped to 8 TB drives. But to realize any size gains, I had to replace both parity drives and 1 data drive, so my first upgrade was 3x 8 TB drives, and that only netted me 5 TB in additional space. It's a unique requirement and that first size upgrade always feels like a waste of money - so again if you can start off bigger in the beginning, it will actually be more economical in the long run. I think this probably applies to most NAS, but I believe this "upgrade penalty effect" seems more pronounced on Unraid vs. other platforms where you can slide in just 1 larger drive and at least get a little more space.

Just to say that I found the Server Layout plugin.

I know the original server layout plugin I used is no longer compliant, and I seem to recall the replacement server layout plugin I use is also no longer compliant. I think my installs are grandfathered in. The plugin I'm using now is called ServerLayout (sounds like what you found) and it is super slow and painful to use now, taking several minutes to open on my server. It's painful enough that I've failed to keep the assignments up to date - I just checked it and at least 5 drive bays are not correctly assigned in my config.

The old one was much better, but I don't think it works anymore. Seems like there's an opportunity for a new solution in this arena. I offered up a suggestion many years back to have the drives color coded according to temperature, so I could see a heat map of my enclosure. People liked the idea but told me to program it myself... so that didn't happen.

For now, I've repurposed my old HTPC as an Unraid server (I took the 1080ti out and replaced it with an old 7600GS). This old HTPC had a decent Antec case and already a 5-in-3 bay. It's a 3770K, 16GB RAM, Asus P8P-Pro motherboard. I installed an Adaptec SATA card with 16 ports, and I managed to fit 12 HDDs (old 4TBs I had in my Synology NAS). The case has lots of ventilation and the drives run cool, so it's not crowded.

Sounds like a good start. If you use a CPU with an iGPU, then you don't need a graphics card at all. I don't recall if the 3770K had an iGPU.

I have 10 data drives and 2 parity drives. I'll add my 1TB Samsung SSD as a cache once I've moved all the data. I love the fact that I was able to build a 48TB NAS with components I had lying around. The only expense was the Adaptec SATA board and the 4 sets of cables... Very satisfying!

Yeah, Unraid is awesome for that. My first server was a bit like that. I know what comes next.

Besides the media duties, I plan to run a VM with MyMovies and a single CCC instance, as discussed with Paul, as the UNraid server will always be on if I'm doing anything MyMovies related. I also plan to use it as a bridge so that I can always control my Oppo 203 clone with the MyMovies iOS App.

Yes! But 16GB is on the slim side for VM duty. Also, does that 3770K have Intel VT tech?

I only have two questions for Paul at this stage (if I may)

Yes, you may.

2) Once you've created a Windows 10 VM, how do you actually install Windows / access the VM? I don't mind using Remote Desktop, but how do you know the settings to access the VM before you've even installed Windows?

From the VM management screen, you can attach to any VM via a built-in remote console. You access it by clicking the VM's icon, then select VNC Remote. This allows you to step through the install process starting from BIOS (well, virtual BIOS), so you can install Linux or Windows or whatever you want. This is more like an IP Management solution, allowing you to see and interact with the console even before there's an OS running.

But once you get Windows fully installed, you'll want to enable the Remote Desktop functionality and switch to using that for the superior experience.

image.png (98.84 KiB) Viewed 2179 times

I'm also setting up the Unraid server so that it goes to sleep with S3, and I plan to use WOL to wake it up when necessary. I'll let you know if that works or not, but it should. I've set the HDDs to go to sleep after 15 seconds.

Back in the early days, I use to S3 my Unraid server. Different Unraid versions occasionally broke this capability and users would have to find new solutions. I haven't tried sleeping Unraid in years, so I'm out of the loop. Since my home security cameras and HDHomeRun recordings are all writing to my Unraid array, I have to leave it up all the time these days. I think there may be some challenges with running a VM inside Unraid and sleeping Unraid, but I don't have any wisdom to share beyond expect a rough road ahead. I'm certainly interested in hearing how you make out in this regard.

I've set up a test array, it took a bit less than 9 hours to build the 2 parity drives, I'm going to run some speed tests (test performance with/without parity) and various reliability tests before I fill up the array.

That's not too bad. Just a note, your monthly parity check speed is directly affected by your mix of drives, if you care about such things. Mine typically runs 17.5 hours with a mix of 8TB and 3TB drives. If I ever complete my migration to all 8 TB drives, it should drop below 12 hours, as those 3TB drives are slowing down the process a lot. The more different drive sizes you mix, the worse parity checks and rebuilds get. Years ago I cared about this more, but these days I've learned to just ignore it unless there's errors.

From a movie copying/watching perspective, performance is adequate for the job. It will never be faster than your slowest drive being read or written, and because it has to read and rewrite parity data, writes are typically half the speed of reads. You will not see gains in performance by adding more drives, unlike RAID 5 type setups, in fact adding more drives usually results in reduced performance.

For maximum performance, you would want to plan ahead and have all identical large, fast drives. But the gains to be had here are probably not worth abandoning using older, slower, smaller drives. This really only matters if you are truly starting from scratch with no legacy drives to re-use. Unraid's ability to swallow up drives of mix-matched sizes is truly awesome, but often users are surprised at the livable performance penalties this path brings.

I didn't run a pre-clean on the drives because they were "pre-prepped" using WD Diag (write zero on all discs from my Windows PC, which I did when I de-commissioned them). They all passed that test, so I didn't feel like running a pre-clean as well, but maybe I should zap the array, pre-clean all the disks and rebuild from scratch.

For a brand new build using pre-owned drives, doing a pre-clear has little to no benefit, so you're fine there. Actually, when you first started up the array and it built itself and parity, that effectively did what a pre-clear does. All zeroes were written to the drives again, and so they were prepped by Unraid to be ready for use, and were given a workout. The pre-prep you did actually did nothing for you, other than alerting you to a bad drive if you had one.

Now that you are up and running, doing pre-clears will actually make a huge difference. To best understand the difference, here's the workflow both without and with a pre-clear:

A) No Pre-Clear: You stop the array, add a new drive, then start the array. Unraid recognizes the drive as new, and so it begins a zeroing process on that drive. Your entire array is in a state of reduced performance/redundancy during this step, which lasts until that drive has had all zeroes written to it (i.e. 5 hours, 11 hours, 20 hours, whatever - depends on the drive speed and size).

B1) Pre-Clear: While your array is still running at 100%, you pre-clear a new drive outside the array. This pre-clear normally takes 3x longer than a non-preclear, as the drive is pre-read, then zeroes written, then re-read to verify. So if you have a drive that would have taken 11 hours in the A) option above, then it will take closer to 33 hours. Some see this as a negative (and for that reason to can turn off extra reads), but many see this as a positive - you are giving the drive a thorough workout, and since most drive failures occur when the drive is brand new, you can weed out the weak ones before adding them to your array. Some users will even add extra cycles to their pre-clear routine, to make the workout last even longer, for a bigger stress test. When the pre-clear is done, if the drive passed then an Unraid "drive-cleared" signature is written to the drive.

B2) You then stop the array, add the pre-cleared drive, then start the array. Unraid recognizes the drive as new but already cleared and ready to use. After a few seconds your array is back to 100%, full performance, full redundancy, without hours of clearing to deal with.

So the advantages or pre-clearing is two fold - weed out weak drives, and reduce the amount of time your array has reduced performance/protection when adding new drives.

I plan to get a 24-bay case (they are out of stock at the moment) and create another UnRaid server with an almost identical motherboard I already have that uses an i5 2500K and also 16GB of RAM.

Since you're planning on running a Windows VM, this feels like a step in the wrong direction. My concerns about your 3770K + 16GB memory are only magnafied here. You probably need to plan on at least 24GB of memory (probably easier to hit 32GB), and I don't know if the 2500K has Intel VT support (required for VM's). While a 2500K was an awesome processor back in the day, it might be on the weak side for what you want to do. Remember, you're running a Win VM inside of a running Unraid server - you need sufficient resources for both. You'll probably also find that you want to run some handy Dockers, and maybe additional VM's. While CCC is light-weight, unfortunately My Movies + SQL Server are demanding, so I have concerns you are under-provisioned in your planning.

I don't want to shy you away from at least trying out your ideas/plans - just be prepared to go a bit bigger.

I'm running an 8-core Ryzen 1800X with 64GB of RAM, which is overkill for what you need, and I have 2 Dockers and 15 VM's - most of them off but in some enterprise software testing I've run 6 VM's at the same time. The point here is that you don't need expensive hardware, but more RAM, more cores, and having VT support are all things you should be planning for.

Ultimately I want to get rid of the non-Unraid servers (or keep them as backups so that I don't lose anything if they die) but it will take a while as I don't really want to buy 24x12TB drives right now...

Yes. This. Many times yes. If you already have it, then keep them as a backup. While a NAS server has redundancy, that should never be confused with having a backup.

I didn't have an old array to use as a backup, so I picked up 5x 16 TB USB drives, and when not in use they are powered off. I power them on about 4x a year, and I use Unassigned Devices to connect them together as a large 80TB drive. I then use a sync job to back up all my changes. These drives cost me about $300 each, so for $1500 I have a true offline backup of my entire movie collection. Soon I will add a 6th drive, but I'm not quite there yet. By keeping them powered off, they are safe from ransomware too. The whole stack is small enough to be portable, so I could store them offsite if I wanted to.

Never trust any NAS to be its own backup.


Hopefully I hit all your questions. I love the topic, so feel free to ask more.

[Manni]

Hi Paul,

Thanks for all the details.

Yes I know that NAS is no backup, I have a proper backup system for my essential files (about 10TB) with offsite storage, but everything else is my DVD/Bluray/4K UHD Bluray collection, so even if I was to entirely lose an array (which is very unlikely), I would only have to re-rip the titles, so that's my backup. Many disks don't keep that well and can't be ripped after a few years, but if I come across a defective disc I simply rent it, copy it, and don't feel bad about that as I have the original. Still, I'd like to have a backup of my most important titles (mainly my 4K UHD Bluray), so that's what the old NAS Boxes will be when I have all my shares on one or two UnRAID boxes.

There must have been a misunderstanding regarding the second UnRAID server. I don't mind if one of them can't run VMs. It will be exclusively used for data, no VMs, no dockers. It might make sense to move the 3770K to the bigger server, but the 2500K should be fine for data only duties. I only need one server to be on permanently, and it might make sense for it to be the smaller one rather than the larger one. We'll see.

I'm going to see how the Win10 VM behaves with 8GB of RAM and one or two cores and an SSD. If I need more RAM, I'll upgrade to 32GB, that's no problem.

Re the 12x4TB disks, they are old disks that I replaced a while ago. They were in the DS2411 (or the DS1211 extension). They were sitting unused since (after I ran WD Diag), so it was a good opportunity to put them back in service. I prefer to have all the drives of the same size.

Even before UnRAID, I always validate disks when I buy them to give them a workout (write zeros on all disks) before putting them into the array, so a pre-check sounds good to me, even if it takes longer. I also always validate disks when I take them out of the array, so that I know they are good and can be re-used when necessary. This is why the 12x4TB were already pre-prepped.

I'm with you regarding less drives of higher capacity, but I've never never been in a situation where I could start from scratch. I've always upgraded whatever disks I had bought initially (as being the best option at the time) to the maximum capacity when it became time to replace them (so 6TB for all my current boxes). I've recently upgraded all the QNAP and Synology boxes to 6TBs drives, so I'm not going to do it again from scratch.

However, when/if I build the new UnRAID server, I could start from scratch with the largest possible drives, and gradually build it. It just becomes quite expensive, for no added benefit from an enjoyment point of view (compared to a new projector or a new AVR), so I try to balance all this.

Currently, I have 2 x 72 TB in the Synology and 2 x 48TB in the two QNAPs. I have some free space, but that's where most of my data lies. The UnRAID with the old 12x4TB is made from unused components (old HTPC, unused drives).

Once I move the content of the TS859 to the Unraid server, the data will stay there (it will be the same size, just 12x4TB). Then the array from the dead TS809 will be mounted on the TS859 (the TS859 disks still containing a backup of the data moved to the unRAID server), and I'll leave them there too.

The second UnRAID server, I have to decide when would be the best time to build it, and what its purpose would be. Probably a consolidation of the data currently on the DS2411/DS1211 and the TS809 data mounted into the TS859.

I have to get S3 to work, there is no way I'm going to leave a server like this running 24/7 unless all the discs spin down when unused and the server itself switches off until it's needed. I have no use for the Windows VM outside of MyMovies/CMC duties, so it really only needs to be on when I either organising/ripping or playing titles. It doesn't need much power, as I'll do the actual ripping from my Ryzen monster, using a MyMovies client. The VM literally only has to run MyMovies server and CCC, and occasionally act as a bridge when I control the Oppo. It will be on the UnRAID server that I will hold my recent UHD Blurays, so it will be on 99% of the time whenever I have to do something in relation to organising or playing the collection.

The server layout plugin worked absolutely fine, so not sure why it doesn't work on your server. It does take 5-10 seconds to launch, but I don't need it all the time, so not really an issue.

I have done some performance/stability tests, here are the results:

I first copied a 55GB UHD Bluray folder from the TS859 to the SSD on my Ryzen monster, to get a reference and make sure that read speed wasn't a factor. The speed was 112 MB/s, which is the top end of my gigabit connection.

I then copied the same folder from my Ryzen monster to my UnRAID share. Without parity, the speed was about the same, 114MB/s (that's with the share empty, it would probably go down as each disk fills up, then up again with the next disk). With 2 parity discs, and the speed dropped to 60MB/s, as expected. This is a significant drop in performance compared to my RAID6 arrays on the QNAP and Synology boxes, but I plan to use an SSD pool once I've moved all the existing data, so when ripping it should actually be the same or faster.

I then tried to move data going through my Ryzen monster (from the TS859 to the UnRAID server). Without parity, the performance was identical (112 MB/s), so that's great.

I tried to use Krusader, but the performance was horrific compared to using my Ryzen Monster, so I will probably use the Monster to move large chunks at top speed during the day, and use the Krusader to move slowly more data at night (II have to switch the Monster off overnight).

I also had some weird issue the first time I used Krusader. It hang with the CPU used at 99% during a transfer, at about 25%, and there was no way to stop the array or reboot. The Adaptec started to sound an alarm beep, which it shouldn't do as it was disabled in the BIOS. I had to hard power off. It made me think and I realised that I had left the Adaptec to the default RAID mode. I switched it to HBA, hopefully it will stop it from monitoring parity, and hopefully it won't cause any more issue if it was responsible. This didn't make any difference re performance, but it allowed the test transfer to complete (after I rebuilt the array without parity).

Still testing for stability, but hopefully I'll be able to start moving data soon if I don't run into further issues.

I'll make an attempt with the VM, thank you for the info, and I'll come back for a final report (and hopefully no furthr questions).

Thanks again for all your help, and for all the information you've put into the thread. Much appreciated

[Pauven]

There must have been a misunderstanding regarding the second UnRAID server. I don't mind if one of them can't run VMs. It will be exclusively used for data, no VMs, no dockers. It might make sense to move the 3770K to the bigger server, but the 2500K should be fine for data only duties. I only need one server to be on permanently, and it might make sense for it to be the smaller one rather than the larger one. We'll see.

Yep, misunderstanding. I didn't realize you were already talking about having 2 servers. I thought you were repurposing hardware, just moving your service license around.

Which happens to be one of my favorite features of Unraid. Your key is tied to the USB stick (so get a good one), and the software is mostly hardware agnostic. I've done MB upgrades where I rip out the old MB/CPU/RAM/GPU, and put in something completely different, and it simply doesn't care. And since the VM's are running on an abstraction layer, as long as you're not passing through hardware then they don't care either.

Try swapping an Intel platform for an AMD platform on a running Windows install - it's a recipe for failure. But on Unraid, success!

I'm going to see how the Win10 VM behaves with 8GB of RAM and one or two cores and an SSD. If I need more RAM, I'll upgrade to 32GB, that's no problem.

For day to day use, 8 GB is marginal for My Movies. But major My Movies upgrades will probably fail - mine did until I threw more memory at it. I think all the DB scripts during an upgrade require more memory than usual. I was running 8 GB, and jumped up to 16 GB to get the upgrades to work. It looked like it only needed a little extra, maybe 10 - 12 GB, and you can divvy it up in Unraid like that if you want which is nice. It wasn't fun troubleshooting this issue, which is why I'm making a point of saying 8 GB doesn't seem to be enough for My Movies. And that's a basic Win10 64-bit Pro install, no other software. Don't be lulled into complacency when you see it appears to work for day to day use.

I also always validate disks when I take them out of the array, so that I know they are good and can be re-used when necessary.

That's a good practice. I've got a small stack of 3 TB drives I retired, and I precleared them after they were successfully replaced, and they are in a stack for emergency replacements of other 3TB drives.

It's also a good practice to have 1 or 2 pre-cleared spares of your largest capacity drive. I always keep at least two 8 TB spares pre-cleared and ready to go. In an emergency rebuild, the pre-clear doesn't offer any time savings, but at least I know the drive is good. You can replace any failed drive with a drive of same or larger dimensions, but you can't downsize.

I have to get S3 to work, there is no way I'm going to leave a server like this running 24/7 unless all the discs spin down when unused and the server itself switches off until it's needed.

The discs do spin down when not in use, and also only the disc being accessed has to spin up. So if you're playing a movie, only the one disc will spin up. That's one of my favorite features. It's also one reason I get so pissed at Binnerup when the API export spins up all my discs for a folder scan I don't need. Depending upon your hardware and controller choices, an Unraid server can end up extremely energy efficient when idle.

I made one mistake in my build years ago, I chose a high performance 24-port HDD controller, not knowing it would consume 55 watts on its own, even when idle. I couldn't believe a controller board could consume so much. My energy efficient CPU was only consuming 15 watts in the same build. Argh!

So if you had a very energy efficient MB+CPU, used on-board HDD controllers, and just a handful of large capacity drives, then you could probably have a server in the 10-20W class at idle, but you would have to carefully make your hardware selections. When shopping MB, know that those that have lots of SATA ports often have extra energy consuming controllers on-board. Just because they come pre-installed on the MB doesn't mean they don't waste electricity. You want to stay within the confines of the CPU/NB/SB SATA ports, otherwise you're looking at energy wasting SATA controller chips. I don't know why they are so inefficient, but they are and the manufacturers don't care. From that perspective, add-in boards are pretty much the same as far as energy consumption.

My 24-bay server case, on the other hand, can consume 30w just from the fans blowing.

Depending upon how often you want to power down Unraid, worst case you can just turn it off instead of S3. Resuming from S3 might be a touch faster, but if you're on the road that might not matter (unless you want to power it on and access it remotely).

If you're truly looking to have two servers, one on full-time for VM duty, then I would put the more VM capable hardware in that build. More cores & more memory, but less storage. You'll probably find that it's all the HDD's and controllers and cooling that consume the most electricity, so your big storage server should be the one you sleep.

I tried to use Krusader, but the performance was horrific compared to using my Ryzen Monster

I know some users have done some testing to find the fastest solution possible. It might even be in this thread. I don't recall now what the result was. I've not experienced that Krusader issue you mentioned, but I also haven't used it in years.

A handy performance tip for you: don't turn on parity until after you've completed your initial data migration. Writes with parity are slower, and add more wear to those drives. You migration will go faster without parity, and since you still have your source data you are at low risk of losing any data. This might make the process a little less painful.

[Manni]

Thanks Paul.

You might have missed it, but I reported the performance with/without parity in my last post. As expected, with parity was twice as slow. That's one of the many pieces of info I got from reading this thread before posting, so I was expecting this. It will definitely save me a lot of time during the migration, so thanks again.

I'll report back when/if I have any news, but it will likely take me a couple of weeks.

Have a great holiday!

[Pauven]

There is a special write mode of Unraid that kicks in if all drives are spun up. Instead of reading parity data from the parity drives, recalculating, and writing back the new parity to the same drives, Unraid will read from all other data drives simultaneously, writing data a newly calculated parity data to the parity drives.

Because it only is either reading or writing to any single drive, this eliminates the 50% performance penalty. Write speed will instead be limited to the read or write speed of your slowest drive. Assuming you had all identical drives, write speed would be close to 100%.

I'd personally rather leave drives asleep, so I've never paid any attention to this feature. But it would be useful during a large migration if parity was really slowing you down.

I seem to recall that there is a setting to enable/disable this feature, and I don't recall exactly what the setting does or if it's enabled by default.

[Manni]

Thanks, I'll take a look and I'll see if I can find it, but for now as I have the data on the QNAP anyway as a backup, I don't mind switching parity off as I would also rather have non-necessary drives asleep.

When I switch to production and enable parity, I'll have a 1TB cache, so I think the limitation will be my gigabit network, given that I assume parity will only be written when mover moves the data to the array (as my understanding is that the cache isn't protected, hence not part of the array, hence no parity is written when data is initially written to the cache. I like that as it means that until mover kicks in, in theory all my drives should be off.

I've already written about 2TB overnight using my laptop (around 60MB/s, so same as Krusader but without the hassle). No issue. As soon as this batch is finished, I'll move to the Ryzen monster to move the next batch, at twice the speed (112 MB/s).

I have about 30TB of data to move, so I should be done in a week or so.

I have a few questions for you:

1) At 00:00 yesterday, I noticed that my write speed on the UnRAID server suddenly dropped, and it looks like it was a scheduled read test, so I cancelled it. Do you know where I can disable/change this setting? I was lucky to catch that one before going to bed, but I'd like to make sure that it doesn't happen during my migration, and after that I'd like to set the time at a time when I'm not using the server, so around 2:00am.
2) I've installed (but not enabled) the file integrity plugin. It sounded like an interesting additional layer of protection, but I'm not sure it's safe to use, so I haven't enabled it yet. Any opinion on this tool?
3) Would you by any chance have a list of your installed plugins? I'd like to see if there is anything that would be interesting to my use in your list.

Thanks!

[Pauven]

my understanding is that the cache isn't protected

It can be. You can create a RAID1 mirror with two cache drives if you use btrfs. I've long had plans to do this, but never got around to it.

Also, a tip: I use my cache drive for more than just caching. I have a few shares that I keep on the cache drive as I want them to be faster and not spin up any HDD's. For example, I have one share I call "Software" and another I call "Music". They're small enough to live on the cache drive permanently, but that means they're not protected. I run a backup script weekly that syncs any changes in these directories into a backup share inside my array. I really wish there was a Share option to say data is stored on "both", but my backup solution effectively accomplishes that.

Along those same lines, I had a 4TB WD Purple surveillance drive mounted outside my array. It's always spun up, and that where my security camera footage is stored. I've found surveillance drives run cooler, consume less power, and are designed to spin 24/7 non-stop. While my video data is at risk since it's outside the array, there's no way I want my parity drives running 24/7, so to me this was a worthwhile compromise. I use Unassigned Devices to automount my security camera shared folder on startup. If I really wanted to, I could add a second Purple and create a mirror for this data, but I feel the odds of me needing the data are low enough that I'm living dangerously.

1) At 00:00 yesterday, I noticed that my write speed on the UnRAID server suddenly dropped, and it looks like it was a scheduled read test, so I cancelled it.

I have no idea what this is. I checked my schedule, and the only thing that runs at midnight is the SSD Trim job, and that only executes on the 2nd day of the month. I'm running Unraid v6.8.3, as I have some drives incompatible with 6.9.x and newer. Maybe you're running a newer version that has something new? And maybe you installed a plugin that's doing something?

The first step in troubleshooting is to go to Settings > Scheduler, and see if there's anything scheduled for midnight.

2) I've installed (but not enabled) the file integrity plugin. It sounded like an interesting additional layer of protection, but I'm not sure it's safe to use, so I haven't enabled it yet. Any opinion on this tool?

Sounds like a good plugin. I've read about it before, but I have no experience with it. The concept is solid. If you click the ? button on the plugin, it will take you to the support page in the forum. I'd definitely recommend reading up on it before trying it (or any other plugins you're unsure about). For example, at the top of this support page is this:

image.png (15.79 KiB) Viewed 2159 times

My Plugins:

image.png (367.86 KiB) Viewed 2159 times

image.png (315.2 KiB) Viewed 2159 times

Many of those come pre-installed these days, but I don't recall which ones. Unassigned Devices + Plus addon should really be stock Unraid capabilities, but at least we have plugins so I can't complain too much. Dynamix Cache Directories is a must have to prevent HDD spin-up just from browsing folders. Nerd Tools is awesome on so many levels. Open Files and File Activity are handy when you're trying to figure out why the server won't shutdown because files are in use, and can also alert you to devices on your network scanning your data (i.e. My Movies, hahaaa). Preclear Disks of course, but I use it with Joe's script, I don't like the stock one gfjardim created (he tries to make it faster, Joe's is slower on purpose). Tips and Tweaks is helpful, especially when you get alerts about new issues moving to new Unraid versions. Nvidia I can uninstall, as I'm not using it anymore - I was using it to pass through a RTX 2080 Super into a Docker for Folding at Home last year. The idea of Dynamix Auto Fan Control is awesome, but somehow it only works right for me after a fresh boot. Pretty much all of the Dynamix stuff by Bergware is great. Hmm, after going through the full list, looks like I like and use everything, except for the Nvidia plugin.

For Dockers I'm using Shinobi_Pro to record my security cameras, and binhex-minidlna which allows me to set up a folder as a DLNA share on my network - this is required by HDHomeRun to record TV shows to a network device. There's hundreds or thousands of awesome dockers, you just don't know what's there until you browse. I've thought about creating a docker to do some CMC stuff, i.e. how about a CCC docker. But that probably means I need to migrate my code to linux, and I would have to pay thousands more for an enterprise license to get that support, so this seems unlikely for the handful of CMC users with Unraid (though the list grows...).

The only problem with all of these great features is that eventually you'll find something that makes you want to keep Unraid running 24/7.

[Pauven]

Thanks, I'll take a look and I'll see if I can find it, but for now as I have the data on the QNAP anyway as a backup, I don't mind switching parity off as I would also rather have non-necessary drives asleep.

Found it under Disk Settings:

image.png (107.73 KiB) Viewed 2159 times

The default is Auto, which I think means if all disks happen to be spinning then it will use the faster method, but I just read the help on it and it states that Auto selects "read/modify/write", which is the slow way. It's possible Lime Tech changed the way this works, and I might be out of the loop, but I'm pretty sure in older versions that Auto would switch to "reconstruct" mode if all drives were already spinning. I've never tested this, but should be easy enough to test.

If you change it to "reconstruct write", then ALL writes will force all drives to spin up for fastest possible performance.

The nice thing is it's a toggle. Use it when you need it.