Home FreeNAS Plex Build Feedback

jasonsansone

Explorer
Joined
Jul 18, 2019
Messages
79
I would appreciate feedback on best pool design for my new build. System will run FreeNAS, Plex, Sonarr, and Radarr (Download clients for Sonarr/Radarr arent on this box). I won’t be using this particular build for NFS, iSCSI, ESXi, ZoneMinder, etc. Data transfers will be almost exclusively by SMB or directly out of the box via Plex streams.

Fixed Specs:
Chassis: SuperMicro CSE-864
Motherboard: SuperMicro X9DRi-F
RAM: 16 x 8GB 1033mhz ECC (128GB total)
NIC: NC560SFP+ for 10Gbe and Intel i350 for 1Gbe
Hard Drives: Shucked WD white label 5400 rpm 10TB

Questions:

CPU: I have available matched pairs of E5-2680 v2 and E5-2697 v2. One set will go in this build, another goes in a separate machine for preparing and transcoding media. Which CPU do you advise for FreeNAS? I am inclined to put the “weaker” 2680’s in the NAS. I prefer to save the 2697’s for x265 re-encoding. Thoughts?

ZIL/SLOG/L2ARC: My research and experience says there would be no value from L2ARC or ZIL SLOG. Can someone please confirm?

vdev and pool design: I currently have a machine with 8 x 10TB in RAID 6 that needs to be migrated to the new FreeNAS build. I will then destroy that array and add the 8 drives to FreeNAS. Data to be migrated is ~34TB. I want to be able to expand over time to fill all 24 bays of the chassis and keep adding to external enclosures if the need arises. I will just keep adding identical vdev’s to the same pool as I need to increase capacity.

Possible vdev Architectures:
3x 10TB RAIDZ = 20TB usable, 6 drive minimum to complete migration, 160TB max chassis capacity
6x 10TB RAIDZ2 = 40TB usable, 6 drive minimum to complete migration, 160TB max chassis capacity

What are the recommendations? At full capacity, RAIDZ would result in 8 pools vs RAIDZ2 being four, thus double the IOPS. However, there is greater parity with RAIDZ2. More smaller vdevs or more parity? What is best practice here? Usable capacity is the same, but I can scale up more granular with only 3 drives vs 6 as I need to add space. I am inclined to select RAIDZ2 for safety, but I wanted feedback.

Thank you!
 

Jailer

Not strong, but bad
Joined
Sep 12, 2014
Messages
4,977
I am inclined to select RAIDZ2 for safety, but I wanted feedback.
There's your answer right there. If you care at all about your data, which you apparently do, then avoid RAIDZ.

How many drives do you have to start with? If it's 8 then I would start there with an 8 drive RAIDZ2 pool and just add another 8 drive RAIDZ2 vdev at a time to your pool. After you add 2 more of them that will fill your 24 bay chassis. IOPS shouldn't be a concern with your intended usage scenario you've described.

My research and experience says there would be no value from L2ARC or ZIL SLOG. Can someone please confirm?
Again with your described usage scenario this would be correct.

I am inclined to put the “weaker” 2680’s in the NAS. I prefer to save the 2697’s for x265 re-encoding. Thoughts?
That would be my inclination as well.
 

jasonsansone

Explorer
Joined
Jul 18, 2019
Messages
79
@Jailer thank you very much. I appreciate you taking your time to assist me.

How many drives do you have to start with? If it's 8 then I would start there with an 8 drive RAIDZ2 pool and just add another 8 drive RAIDZ2 vdev at a time to your pool. After you add 2 more of them that will fill your 24 bay chassis. IOPS shouldn't be a concern with your intended usage scenario you've described.

I have one drive that was a hot spare and will be purchasing additional. I may steal one drive out of the existing RAID6 array (I realize the existing array would be in a degraded state... plan would be to transfer data out ASAP via 10Gbe). I had thought about doing 4 x 6 but 3 x 8 would ultimately yield more usable capacity. If been watching the 10TB Easystore prices waiting on Black Friday specials.

My storage will be mostly WORM. Once a video file is in its final, encoded state it isn't going to ever change. Ignoring the IOPS, what would be the difference in resilver times or other performance metrics for 4 vdevs of 6 drives versus 3 vdevs of 8 drives? I know I have even more parity but less usable space.
 

Jailer

Not strong, but bad
Joined
Sep 12, 2014
Messages
4,977
Ignoring the IOPS, what would be the difference in resilver times or other performance metrics for 4 vdevs of 6 drives versus 3 vdevs of 8 drives?
I couldn't tell you for sure. Resilver times vary depending on drive size, number of drives and the amount of data you have on them. I currently have about 8.8TB used of a 6x8TB pool and my scrubs take about 5 hours.
 

jasonsansone

Explorer
Joined
Jul 18, 2019
Messages
79
I have been running this build for several months and am generally very pleased. However, I need the communities advice on how to solve a few performance issues. The actual build is as follows:

Chassis: SuperMicro CSE-864
Motherboard: SuperMicro X9DRi-F
CPU: 2x Intel E5-2697 v2
RAM: 8 x 16GB 1866mhz ECC (128GB total, expandable up to 256GB)
NIC: NC560SFP+ for 10Gbe and Intel i350 for 1Gbe
Main Pool Config: 2x RAIDZ2 vdevs each consisting of 6x10tb 5400RPM
OS: FreeNAS 11.2-U7
UPS: Each of the chassis's redundant power supplies is fed by a separate, redundant APC Smart-UPS 2200 RT on their own dedicated 20amp circuits

I have separate jails for Plex, Sonarr, Radarr, Lidarr, Jackett, NZBGet and qBittorrent. I have separate headless machines I use to ingest / reencode media. They access FreeNAS via SMB over 10GBe NICs. Writing large files isn't an issue and can sustain ~550MBs.

My problem is when I attempt to create remuxes from disk ISO's over the SMB share. Speeds bog down to ~70MBs. It is far worse if there is anything being downloaded in NZBGet. If NZBGet is saturating my WAN at 100MBs, speeds in MakeMKV drop below 20MBs.

I have a spare 1TB Samsung 970 Pro that is unused at the moment. Will adding it as L2ARC help anything in this build? What else, other than adding 2 more vdevs to populate all 24 drive bays on the chassis, help? I am essentially starved for IOPS, but I don't know the best way to tackle the problem.
 
Top