I have been running TrueNas Scale successfully for about a year now.
My installation is as follows:
4x WD Red 4 TB Drives in RAID Z1 (Data) 1x Samsung 2TB NVME (boot Pool). 1x Samsung 500GB SSD (striped, for caching, temp storage etc.) 1x Silicon Power 2TB NVME (not installed yet, but would like to use)
CPU: Intel I7-9700K @3.60Ghz
My primary uses for this NAS are:
Data Storage/Sharing
Media Server (Plex and/or JellyFin)
Areas I’d like to improve:
Since I am running media servers, transcoding seems to be inevitable and imposes quite a hit on the CPU. So, I’m hoping to give the media servers more “umph” in the transcoding and metadata creation by moving those I/O operations over to NVME.
I was thinking of creaating another RAID array that be made up of the 2, 2TB NVME drives.
Currently, my Boot pool is just a single NVME and I doubt if I actually need to allocate all that space for boot purposes. So, I was thinking about reorganizing things a bit.
New Proposed Configuration
Move Boot Pool to the 500GB SSD
Create an array for the 2TB NVME
This will leave the configuration looking like this:
4x WD Red 4 TB Drives in RAID Z1 (Data)
1x App Pool consistong of 1x Samsung 2TB NVME, and the 2TB Silicon Power NVME.
1x Samsung 500GB SSD (Boot Pool)
Is this possible to do without a full re-installation? Does it looks like a good idea?
I’d argue that since you’re re-working all of this just for transcoding to just add an Intel Arc A380 & pass it through for Plex/Jellyfin. You’ll get a cheap beast for transcoding.
This’d cost (for me) $10 less than the 2TB nvme you quoted. It would also generally require no additional work than slapping it into the system & making it available to Plex (or Jellyfin).
To add - you can have Plex transcode to memory which would be faster in just about any way while you’re streaming:
I think the easiest and simplist is to export your Raid-Z1 pool and do a complete reinstall of TrueNAS, import your Raid-Z1 pool, create your mirrored NVME and then set up your apps all over.
Install Truenas to the 500GB SSD. import config (same location as pictures above, but you’d select “import” instead.) This sounds scary, but honestly takes 2 minutes - you won’t lose data on any of your other pools doing this unless you select the wrong drive 5 times in a row during installation.
For 2TB NVMEs, creat pool, mirror, select both. Move your apps to it.
Good idea. The problem is, I am using a Mini-ITX board. The small form-factor wouldn’t allow me to add the ARC b/c the one PCI slot I have is already being used! Great idea though. I am using the compression that comes with the CPU. Not as good as the ARC but it’s something.
Ah The iGPU can’t keep up? Just to confirm, you did pass it through to Plex/Jellyfin? That should take most the load off of the CPU itself as it should still be very serviceable for transcoding.
I have the passthrough enabled. I haven’t done an A/B comparison to see how much it helps. I did do a test with a copy of 2001 - A Space Odyssey (4K+HDR) it pauses every minute or so for about 4 seconds. That’s using Plex. I think JellyFin behaves nearly the same.
Yeah - HDR is a killer. Tone mapping is where iGPU bites the dust entirely.
What motherboard do you have & any kind of slots open on it? There could be room for some sketchy stuff that I wouldn’t recommend for anything other than getting some transcoding going since that is likely the least mission critical part of the NAS.
Now THAT is interesting… Considering you ain’t using wifi (I hope), having that little M.2 slot makes me wonder about doing something sketchy like this: https://www.newegg.ca/p/17Z-0123-00001
Having 4 lanes of pcie for transcoding ain’t the end of the world & is arguable on if/at how many streams it would noticeably impact performance. Having the riser be powered is important since the m.2 slot ain’t going to supply enough power that the GPU might expect from PCIE. Might make for an interesting idea depending on your risk tolerance.
…Of course this’d mean hotrodding your likely beautiful case with a dremel or angle grinder & then figuring out some structural support for the GPU. Not everyone’s cup of tea. Or you then pair it with an x16 to x16 riser (which for pcie gen 3 specifically should be fine) to avoid cutting into the case.
This idea would also disable the following sata port:
Edit:
Anyway the only reasons I’ve coming up with these ‘ideas’ - is because frankly you can optimize storage as much as you want, it ain’t going to give you the transcoding performance improvement you want. Nothing wrong with changing you boot drives & mirroring your m.2s (those are good ideas, you should do them) which is why I gave quick instructions in an earlier post, but it ain’t gonna noticeably improve transcoding.
…a third, likely, less stupid option would be to run plex off of a completely separate machine with a dedicated GPU (since you got HDR content) else & give it access to the media on your NAS.
If these are actually WD Red drives (as opposed to WD Red Plus or Pro) then they are SMR drives and entirely unsuitable for use in a redundant ZFS pool. However early WD Red drives were CMR and were later rebranded WD Red Plus, so you need to check using the exact model number.
Using even a 500GB drive for a boot pool is crazy if you have another use for it - but if you don’t mind wasting 90%+ of the space, then it should be fine.
I am assuming that you have a 2nd native M.2 PCIe slot to put the second NVMe card into. If not then you need to provide details because it may not be such a great idea to mirror the existing NVMe card, and instead better to replicate it to HDD as a backup.
IME/IMO this is way more power than you need except perhaps for real-time Plex transcoding (unless you are running VMs or heavier apps than Plex). I do my transcoding when I create the media and as a background task, and my Intel Celeron J3355 Dual-Core 2.0 GHz CPU can transcode a single file 1080p file to 720p at between 0.7x and 1.5x speeds. Despite such a low power CPU, if I watched TV for 12 hours a day I literally still could not watch media faster than it can transcode it.
But obviously if you already have the CPU, then more power is better than less.
You don’t say what your memory size is - but if it is 16GB or more, then for Plex use you are absolutely good to go. (Mine is only 10GB and I still get 99%+ cache hit.)
Great idea. Store the media files themselves on HDD, together with any backup data. Store shared data on HDD or SSD as your performance needs dictate. Store all Plex metadata on NVMe SSD. If you have enough memory, do Plex transcoding in memory and if not on NVMe SSD.
This is what I do on my underpowered, under-memory system and it still performs brilliantly.
Yes - it looks like a great idea to me.
I am guessing that what you really want to do is to avoid rebuilding your existing RAIDZ1 HDD pool because that would be a big pain, and yes you can do this rebuild without needing to offload your data and rebuild the HDD pool.
However it will almost certainly be easier / quicker to reinstall TrueNAS on your 500GB drive and import your configuration than to try to move your boot drive from a 2TB NVMe to a 500GB SSD. (If your existing boot drive has been installed using default install, it will be 2TB in size - if you previously did the hack to install TrueNAS to the NVMe using <= 500GB leaving the rest spare, then you could try mirroring the boot drive and then removing the NVMe from the mirror, but it might still be easier just to reinstall.)
Here are the steps I would use:
Remove all use of the 500GB SSD from TrueNAS - no use as cache or temp storage or ZFS pools.
Backup your system configuration file.
Backup your system configuration file (again).
Reinstall the same version of TrueNAS onto the 500GB SSD and boot from it.
Import your backed up configuration file.
Clean the existing 2TB NVME and create your apps-pool.
You are good to go. (You may need to migrate your existing Plex metadata from HDD to NVMe - ask if you need help with this.)
P.S. Make sure that you run scrubs, SMART short and long tests on all drives. If you haven’t implemented @joeschmuck’s Multi-Report script then you should do so.
Mind that a WiFi M.2 key E port is typically PCIe x1 and USB, not your usual M.2 key M.
What may be done, however is x8x4x4 bifurcation of the x16 slot.
Thanks, I think I’m going to give this a try! One thing though (and this may seem lame), how to you “Reinstall the same version of TrueNas onto to 500GB SSD”?
Let me know if this is correct:
Download the correct version of the TrueNas Scale ISO.
Burn the ISO to thumb drive
Boot from thumb drive.
When asked where to install TrueNas I would select the 500GB SSD
Note sure when to reload the configs, but I assume I’d need to do it though the web UI once TrueNas is again available.
Follow remaining step (6) to see up the Apps Pool on NVME
Yes - you can download the configuration file to your PC from the existing UI using your browser, and upload it to the new UI from your browser.
My advice would be to research how to restrict the boot-pool size when you install TrueNAS and set it to (say) 128GB (which is still way bigger than the 16GB required) but smaller than pretty much any SATA SSD drive you might buy to mirror to in the future.
AFAIK, doing this shouldn’t cause any future issues (I have done this and upgraded several times without any problems), and (providing that you don’t do anything with the remaining space other than swap if you decide you need to turn it on) your system should even remain within the support guidelines from iX.
I as able to to successfully more the boot off to the SSD. I found two like NVMEs to use for the App Pool pool and striped them together.
All is well, except now my login to the NAS are throwing “Access is denied” error. I there something happens to the users/creds when restoring from a system backup (the one you create from TrueNas and upload to the NAS after an install).
The iGPU can’t keep up? Just to confirm, you did pass it through to Plex/Jellyfin? That should take most the load off of the CPU itself as it should still be very serviceable for transcoding.
