System Longevity?

[bigzaj]

My current TrueNAS system came together ~3 years ago. Right now it is:

- Dual Xeon E5-2670v2
- 192gb ECC ram
- 134Tb usable pool for Media / Files
- Vdev1 - 10x WD Red 6TB RaidZ2
- Vdev2 - 10x Seagate Exos x16 14TB RaidZ2
- 1tb system pool for Jails / VMs on mirrored SSD
- Chenbro SC846 Case

I'm feeling a bug to upgrade or do something, and also curious just how long this system should last me. The only hiccups I have right now are around Plex Transcoding and inability to have a "Gaming VM" for VR (which might be cheaper to get a dedicated PC here, I get it).

I've been considering an updated build, something that would support storage north of 200TB+ but not sure it is necessary.

Questions:
1. How long should this setup last me from a hardware longevity standpoint (I know variable, but assuming average use and temperature controlled)?
2. Keeping the current system as is, any issue dropping in a video card to handle transcoding when TrueNAS scale hits and offloading transcoding to the GPU?

I've been considering trying to do a dual Epyc build or a new gen Xeon single ... exchanging cores for GPU.

Appreciate any feedback.

 

[hescominsoon]

I have a dell r520, a dell r310 and a dell t20 all in production..they all work perfectly fine. Use the reporting function..look at your metrics...My most recent system is the x20. all aforementioned systems are in my signature. I have several minis i have deployed as well..those are used as backup/replication targets for larger machines.

if you are seeing performance problems then an upgrade might be warranted..when it comes to critical infrastructure like storage my rule is..if its not broken or performing badly don't mess with it.

 

[Heracles]

1. How long should this setup last me from a hardware longevity standpoint (I know variable, but assuming average use and temperature controlled)?

Assuming average use, this system is critically oversized.... 192GB of RAM and 160 TB of storage is nowhere near average use. As such, this system should last for over a century!

As for keeping a system online, I consider the access to replacement parts as an important factor. Power supplies are the most frequent failure. As long as you can get replacement parts, that will help keeping the system going.

Another thing to consider is the server's role. Here, out of my 3 servers, Thanatos is by far the oldest and least powerful. But being an offline server, it just needs to be barely able to run the pool and receive the ZFS data. Should he die one day when I try to power it On for a sync, it will not be dramatic. I will still have my 2 main servers online, the offline copy will remain safe and it will be easy to migrate my offline data to any replacement server I will get my hands on then.

So for you, what is your backup plan ? With such a large pool, it is surely not DVD copies.... If you do not already have 3 servers to enforce the 3 copies rule (see my signature), I would definitely keep that system in the loop and use it as a 2nd or 3rd copy. If you do that, you can migrate some RAM from it to whatever will be your new server if they are compatible.

- Vdev1 - 10x WD Red 6TB RaidZ2
- Vdev2 - 10x Seagate Exos x16 14TB RaidZ2

inability to have a "Gaming VM"

RaidZ vDevs are not very good at IOPS. On the other side, VMs needs more IOPS. Whatever is designated as "gaming" usually refers to a very powerful version of that. As such, a gaming VM would be a very high performance VMs. You need mirrors (and many of them) to ensure high IOPS. Considering you have 20 drives as of now, that means a pool of up to 10 mirrors, 5 times more than your 2. That would definitely boost IOPS but requires to destroy the pool / build a brand new pool.

 

[Constantin]

I like keeping my things simple - the storage server does storage, the nodes in the network do transcoding, gaming, or whatever else. It allows each system to be optimized for its task and in the long run it’s also allowing stuff to kick around much longer than trying centralize it all and then having to throw out the entire core just because some aspect doesn’t run right anymore.

For example, just as my dumb living room TV has been around since 2006, the Mac mini that’s feeding it has been around since 2012. No need to update either until High Sierra no longer allows updates for various streaming services to operate. Then I’ll replace it with the next mini and maybe the display too.

i realize that this approach to things will cost more in terms of hardware / space but on a long term basis I do think I’ll come out ahead. Additionally, it allows the server to operate far from occupied spaces and minimize its power consumption.

 

[Sjöhaga]

I'm in a similar situation. While not urgent I have started to plan the next storage server. My current storage server has WD Red 4tb drives in its oldest pools with power on hours 40k+. The server is based around a Supermicro gen 9 board and all in all it is still enough for my needs.

But gen 9 is getting old, and drives from 2014 is old . What I have done is I have spare parts on the shelf for most things except the mobo itself. And for the things I dont have own spares I know shops that have the rest. So I am confident that I can keep this old system running for several years still, and when I finally do build the new system, as Heracles said - this one will become the backup for the new.

I might even start replacing the old pool with new larger drives before I start on the new build (unless that lotto ticket pays off, then I'll do both)