Plex Media Server Build Questions!

[1kca]

So I'm building a Plex media server; I have all my parts except the storage drives. I have two main questions...

 

1. Ideally, I'd like to stay within Windows. Is there a software solution similar to RAID6 that will work well for a media server in Windows? I'm aware of Windows Storage Spaces which I've attempted to use in the past but it refused to detect all my drives, even though they had all been formatted in Windows that same day. It doesn't have to be a RAID solution, but something similar. I want a giant pool that offers some kind of redundancy in case 2 drives were to fail. 
2. Can I use any drives or should I go for purpose built NAS drives such as WD Reds? I recently bought 8 Seagate IronWolf drives, but they all died when the PC went to sleep while wiping them. I don't necessarily want to spend the extra on NAS drives if I don't have to.
3. Is it worth having a separate SSD as a transcoding directory? 

 

Originally I was trying to decide between: WD Reds, HGST Deskstar NAS drives, Toshiba N300s or Seagate IronWolfs, but I'd just as happily buy WD Blues or whatever I can get my hands on if they're cheaper and will work just as well. 

 

I should say that the PC will be on almost 24/7. I'll turn it off here and there for updates but really, that should be it. 

 

Media Server Specs:

 

CPU: AMD Ryzen 7 1700

Motherboard: MSI B350m Mortar

RAM: Adata XPG 16GB 2400MHz (saw it going for £79.94 and went for it)

GPU: Nvidia GT 710 (no iGPU on Ryzen)

OS Storage: 256GB Samsung 960 Evo NVME SSD

Mass Storage: TBD

PSU: EVGA SuperNova 650 G2

Case: Fractal Node 804

Misc: LSI 9211-8I 8-port HBA Card

 

[kerradeph]

Does it have to be Windows? FreeBSD/FreeNAS has been working pretty well for me. It has ZFS which supports a form of raid 6. Realistically there is the software raid option in Windows disk management but I don't know if that supports raid 6.

 

Drive type doesn't matter. NAS drives are generally cheaper and since Plex is a pretty low load on the drives you don't need much speed. 

 

I don't think transcode would be much of a load on spinning rust if you're just doing live transcode on streams. 

 

Also, make sure you have the HBA card flashed to IT firmware so you don't get any strange things happening. 

 

My Plex setup is the OS/plex is on a pair of mirrored 128GB Samsung SSDs with a pool made of 8 WD 6TB reds in RAID 10. All media is stored on the pool and all logging and workload is on the SSDs.

[1kca]

It doesn't have to be Windows but I'd very much prefer it to be, just because I'm more familiar with it and the fact that it will act as a living room PC for a couple of months before going in a closet somewhere.

 

That's a plus on the drives - means I can just get anything big enough and call it a day. 

 

You wouldn't have a link on how to flash the HBA card would you? Wouldn't be surprised if that's what killed my Seagate IronWolfs

 

Edit: Found a couple of decent guides

Edited June 11, 2017 by 1kca
Google'd

[kerradeph]

Don't get the Seagate archive drives that use shingled layout, they have really strange read write patterns that can lead to annoying read write times.

[1kca]

Not sure what that is.. I'll likely just avoid Seagate's altogether this time around. 

 

Out of curiosity, what made you choose FreeBSD/FreeNAS over another solution?

[kerradeph]

Similarity to what I used at work. I wanted to use it as a training exercise as well and it happens to work fairly well as a media machine. 

[1kca]

Fair enough. I think that pretty much covers it for points 2 and 3, as well as the HBA config I didn't realise I needed lol.

 

Still a bit torn on the storage management solution. I'm going to look into FreeNAS - it comes up way too much to ignore really but I'm still hoping for Windows / Windows compatible solution I can use. 

[brwainer]

1 hour ago, 1kca said:

I'm aware of Windows Storage Spaces which I've attempted to use in the past but it refused to detect all my drives, even though they had all been formatted in Windows that same day.

Don't format drives if you want them to be detected by storage spaces. Open a command prompt then diskpart, "list disk","select disk x" (be very careful to select the right number, diskpart doesn't ever ask for confirmation), then "clean". The disk is wiped of the header info, it won't even be GPT nor MBR anymore. It will be like a brand new unformatted disk. I have had 100% success getting reused drives to be detected by storage spaces this way. The problem usually is that there is a small UEFI partition on the drive that can't be removed in Disk Management, or it is an MBR partitioned drive.

1 hour ago, 1kca said:

I recently bought 8 Seagate IronWolf drives, but they all died when the PC went to sleep while wiping them.

That is something I've never heard before. Did you RMA them with Seagate? I'm sure that's something they'd like to investigate.

1 hour ago, 1kca said:

Is it worth having a separate SSD as a transcoding directory

You don't necessairly need an SSD as a transcoding directory, but it should be at least a different drive then where the media is stored. By default its the C drive.

 

 

All in all I think you should give Storage Spaces another try. You can PM me if you have any issues with it, I'd be glad to help out. My experience with Storage Spaces has been great, and even after setting up a few ZFS pools I still prefer Storage Spaces. 

[ChalkChalkson]

WD Reds and Seagate Iron Wolf drive are very nice to have, but if you already own enough drives, especially if you run RAID 6, go with the old drives, and when one fails, buy a dedicated drive.

 

With the SSD, I would say get one if you can, and just use it as cache and transcoding directory, it makes it quicker to write your films to the array, but 2 SSDs are definitely overkill.

 

And what you might want to consider is running an OS like unraid in the background and using a VM as your "normal" PC, unless you use VR, or play competitive CS GO (one of my friends swears he can notice the difference, I can't). You seem to have plenty of cores and a plex server doesn't need much.

[1kca]

@brwainer

 

Really appreciate the input!

 

I didn't realise pre-formatting the drives would cause an issue. Can't see why it would. The weird bit is that I think 5/8 were detected and added to the pool just fine - it was just the last 3. Tried all sorts and indivisually, they were fine but who knows. I bought them from Amazon so I returned them to Amazon without a fuss but it may have been worth giving Seagate a shout - they may have had some ideas.

Thinking about it now, after what you and kerradeph said, I'm thinking there was more to it. I think it could have been the HBA card not being configured correctly, but even if not, diskpart may have been able to resurrect them. 

I tried them in a dock as well as internally connected to a different PC but none of the drives were detected at all within Windows.. Mystery

 

As for the transcode drive, I'll see how it goes with it on the boot drive. Should be more than enough space but if not, I'll just pick one up down the line. 

 

I'm still going to look into FreeBSD/FreeNAS, but I'll give Storage Spaces another shot first. Makes more sense to try that while Windows is already all set up and then switch over if it doesn't cut it

 

Edited June 11, 2017 by 1kca
Formatting

[1kca]

6 minutes ago, ChalkChalkson said:

WD Reds and Seagate Iron Wolf drive are very nice to have, but if you already own enough drives, especially if you run RAID 6, go with the old drives, and when one fails, buy a dedicated drive.

 

With the SSD, I would say get one if you can, and just use it as cache and transcoding directory, it makes it quicker to write your films to the array, but 2 SSDs are definitely overkill.

 

And what you might want to consider is running an OS like unraid in the background and using a VM as your "normal" PC, unless you use VR, or play competitive CS GO (one of my friends swears he can notice the difference, I can't). You seem to have plenty of cores and a plex server doesn't need much.

Thanks, I actually didn't have the drives before unfortunately. I "had" to buy everything new. 

 

As for writing to the SSD, I think that's a good shout. I'll leave it for the time-being, just because the boot drive I have should have plenty of space as is, but if not, at least I'll know I'm not wasting money by buying a sata SSD for it later on 

 

Also, I already have a dedicated gaming machine which is a plus, so this will eventually be a dedicated media server 

[1kca]

Thanks all for your input - I think I'm set 

[AshleyAshes]

I use Kodi, not Plex, so my server ONLY serves data over SMB, but I could just as easily toss Plex server on it if I wanted to.

 

What I'm using, and what may help you, is FlexRAID:

 

In short, FlexRAID is a paid product and it's a snap shot redundancy system.  You can put in as many storage drives as you want and even add additional storage drives later.  It uses at least one dedicated 'parity' drive that must be at LEAST as large as the largest storage drive.  As you can see, I'm using Seagate 8TB Archive drives as my larger drives and one is used to store the parity data.  As mine is configured, it only updates the parity data every Sunday (Monday, really) at 3am so any data added in the week is unprotected until the scheduled parity update.  The core advantage of this is that this system is highly flexible and there's no limit on storage drives.  Once my system is full I can basically just add yet another 8TB storage drive and the whole thing will have 8TB more storage.  (However it WOULD have to regenerate ALL parity data rather than upst update a snap shot like it does weekly, which for the size of my data set, could take upwards to 2-3 days)  For every parity drive you have, you can withstand one storage drive failure.  If you want to be able to withstand two failures, you have to employ two parity drives.  Also, even if multiple storage drives fail and you don't have the parity, you will still have the data on the operational storage drives.  So even in the 'worst case scenario' where you can't rebuild from parity, you'll only lose as data on the drives that actually fail.

 

Additionally, you can mix drive sizes, again the only condition is that a storage drive can't be larger than the parity drives.  The data is also stored as plain old NTFS data on the individual drives, plus some meta data.  As a result, if your machine ENTIRELY explodes, you can STILL just pull the operational drives, put them into ANY other Windows machine, and copy the files off those drives.  No special tools necessary.

 

This is NOT the best system for 'important data' but for my hoard of media files being able to sustain a single drive failure is 'protection enough' for me.

 

My server runs Windows 10 Pro, BTW

[brwainer]

31 minutes ago, 1kca said:

I didn't realise pre-formatting the drives would cause an issue. Can't see why it would. The weird bit is that I think 5/8 were detected and added to the pool just fine - it was just the last 3.

Pre-formatting as GPT with no volumes (empty partition table) is usually fine, and most of the time having a single partition on the drive is alright as well. But I've had the following conditions be the cause of Storage Spaces not showing a drive:

-non-Windows partition on the drive, e.g. Ext4 - I think this is a safety mechanism, because with NTFS it can check if the drive is empty and OK to wipe

-UEFI partition on the drive, usually 128MB or 512MB - these come from installing an OS onto the drive, or if the drive is the first one connected to the system (in position 0) when an OS is installed, a copy of the bootloader will get put onto the first drive since that's where the bootloader needs to be if the boot order gets reset. Disk Management won't allow you to delete this type

-A partition on the drive that doesn't use 100% of the space, even if only short by a few MBs.

-MBR partitioned drive

clean in diskpart resolves all of these issues, and I got the tip from some of the original Storage Spaces documentation back in the early 2012R2 days when people were testing it in labs before trusting it in production. When testing a system in a lab you are more likely to have all sorts of weird data on your drives then when putting a new system into production, so I think less people talk about issues nowadays.