I don’t have all of my computer parts yet but they are slowly being delivered. As a TrueNAS newb, I’m beginning to think through how to install TrueNAS. I have what I would consider just a general installation for data storage with 8, 6 TB HDDs in RAIDZ2, and using ClamAV for antivirus (I have yet to see a tube video on how to use ClamAV with TrueNAS)…nothing crazy. I’ve watched a lot of tube videos on how to install TrueNAS and have only seen one that is similar to how I would configure the installation. The video is at least one year old though.
I have a motherboard that I will be adding 2, NVMe M.2 SSDs to - 1) 1 TiB and 2) 2 TiB. I will be installing TrueNAS on the 1 TiB NVMe M.2. The steps I thought through are:
Install TrueNAS to 1 TiB M.2 SSD
Storage > Create Pool and select to Encrypt it
Credentials > Local Users > Add a user and related password > make sure Samba Authentication is checked
Datasets > Add Dataset > Share Type > SMB > Save
Shares > Windows (SMB) Shares > Add > Path from /mnt/Pool/select top level directory > Save
Shares > UNIX > ??? [I use Linux primarily but the video didn’t show what to select here]
Datasets > select pool top level directory > Permissions > Edit > Add Item > User > select User from dropdown > Permissions > Full Control > Save Access Control List
Localization > change Timezone
How does the install look?
Now for a question that I don’t have an answer for. I remember watching a tube video where the person stated something to the effect of creating a metadata pool so I don’t lose any data. Hmm, I haven’t been able to find a recent tube video on how to configure for the metadata. Is installing it on the remaining 2 TiB NVMe M.2 a good idea? Would anyone have any steps on how to configure this metadata pool on the 2 TiB NVMe M.2? Or should I not use the 2 TiB NVMe M.2 for metadata and just go with the installation steps above? Remember, I’m just a newb here so I don’t know what I’m getting myself into and I don’t want to mess up the installation on the first try.
Hi - I am a newbie myself, completed my install about a couple of months ago. The biggest advice I can give you is to rely less on YT videos and spend some time reading the documentation to really understand, at the bare minimum, the concepts of vdevs, pools, and datasets, then look at your storage and what you want to do with it, plan ahead how you are going to organize them, before you start. I personally spent several weeks just reading and figuring out the basics before I decided on what to purchase, much less actually installing.
First of all, you need a drive dedicated only to TrueNAS so that is going to be on one of the M.2 - great (know that the overwhelming majority of that storage will be unused but these SSD are super cheap these days so it may be fine).
Then you are left with two 2TiB drives, and two 1TiB drives (the latter two are one SSD and one HDD). So now you need to make some choices as to what your vdev(s) are going to be. With your number of drives we are talking 1 vdev per pool, so for example you could make a mirror vdev out of the two 2TiB drives (which will give you a pool with a bit less than 2TiB capacity) and another mirror with the 2 1TiB drives (which will give you another pool with a bit less than 1TiB capacity). Or, you can put all 4 drives in 1 vdev with RAIDz1 but because they will be used as if they are all 1TiB drives that will give you a single pool with a bit less than 3TiB capacity and a single drive redundancy for the vdev. Those are not the only 2 options.
You did not mention any use cases other than SMB shares so nothing about any apps. I have my apps in a separate pool but because I can more easily recreate the apps pool, I have that one with no redundancy (1-disk vdev).
So there is still some real basic stuff that you should spend time figuring out, way before you get into setting up metadata vdevs, which as I understand, are not really needed/helpful in most cases, so I probably would not start with that as your first approach. Also, RAM is so much more important in terms of performace so make sure you have enough before you start thinking about metadata vdevs. And they absolutely have nothing to do with not losing data.
I agree that reading the documentation would be great, however, I haven’t been able to find a PDF manual for 25.04.1. Any ideas where I can find the PDF so I can print it out instead of sitting at the computer trying to read the monitor?
As far as apps are concerned, I mentioned ClamAV to run antivirus checks on my data.
Ok, so I won’t go down the metadata road…yet. So, I’ll install TrueNAS on the 1 TiB M.2 and I’ll configure the 2 TiB M.2 for fast cache, I think it’s called L2ARC. Hopefully that configuration will go smoothly. Ha!
General rule is that unless you’ve maxed out your ram and your work load is such that you’re getting more ARC misses than hits, you really don’t need a L2ARC. Without a quantifiable usecase & data to back up a need for one; at best L2ARC won’t do much, realistically it will degrade performance.
Metadata cache is dangerous because if it dies, so does your data. It’ll likely be the weakest link without planning for redundancy.
Toss your VMs and Apps on it instead your other nvme imo.
Too many videos make slog/l2arc/specialvdevs look like some magic ‘more hardware, more better’ silver bullet. Even my answer is grossly simplified.
If I’ve learned anything on my time on the forums is that real performance improvements come with understanding of the feature + understanding of your specific workload… or money. The first two are boring, and the third is expensive.
Good luck I’ve basically learned everything from searching the forums.
@Fleshmauler Thanks for the reply. I’m installing 128 Gig of Kingston unbuffered ECC RAM into this build with 8, 6 TB spinning HDDs. Trying to future proof this build for a while if I can.
Yeah, I’m not finding a manual I can download and read either.
Beauty - always happy to see ECC. I strongly doubt you’ll need L2ARC, unless you have an exotic workload or are going for more than just home use.
I mean, there are manuals that you could arguably print - but even then if I had to choose between printing TrueNAS scale documentation vs Oracle’s ZFS Administration, I’d prolly go for Oracle. Won’t include the UI, Apps, Virtualization, or any of the features of TrueNAS… but you’ll be a zfs pro!
For everything else that make TrueNAS Scale awesome in a not really download friendly format:
@Fleshmauler Looking a little further into the metadata question, I found an article called Fusion Pools. Interesting to know your thoughts on this to keep on the 2 TiB M.2.
Sounds like a more complicated version of a meta data special vdev. Same risks; if you have any problems with that singular M.2 - you lose all your data.
Special vdevs do improve the performance of spinning rust (HDDs), but come with heavy risks if you’re not using hardware designed for 99.99999% uptime + redundancy + a controlled environment… or if you’re brave I guess.
Even the article goes into detail about dangers & recommending a UPS if you’re going to try it:
The metadata special VDEV is critical for pool operation and data integrity, so you must protect it with hot spare(s).
UPS Recommendation *expand*
When using SSDs with an internal cache, add uninterruptible power supply (UPS) to the system to help minimize the risk from power loss.
I’m certain that there a benefit using a fusion pool vs just a special vdev, I’m just not familiar with what it’d be. Risks would pretty much identical though.
(disclaimer, I mean this politely) What are you planning to use the NAS for that you’re this worried about having various caches? Honestly eight drives should have no problem getting close to saturating a 10gig link on read performance - write performance and iops should be generally acceptable for SOHO use.
Edit: worst part is once you’ve made a pool with a special vdev, that’s it. There is no going back short of copying all the data to a new location & starting from scratch. So I can’t even recommend you play with it & see if you notice a difference.
@Fleshmauler That’s exactly what I needed to hear. I appreciate your analysis. I’ll use the second 2 TiB M.2 for Apps. Now I just need to figure out how to configure a single M.2 for Apps.
When you got o the storage tab, click create pool, only select that nvme. It’ll warn you that there is no redundancy (you have that NVME die, then your apps are done; but if you set things up correctly, the data that feeds them isn’t; ie keep your movies for Plex on your HDD pool) - you can always add another one later if you feel it is worth the investment.
Then go to Apps tab, choose pool, select the new pool you make on your 1 nvme pool; download, configure, enjoy! You’re about to learn that there is a lot to learn
It’s also advisable to ask for advice before buying.
Totally wasted. Get whatever old small drive you may have lying in a drawer of spare parts, or get some cheap 100-128 GB second hand drive, for boot and put that 1 TB to better use elsewhere.
That would be NFS, but Linux can also use SMB shares.
That is a Very Bad Idea™: If this single drive ever fails, you loose the entire pool.
To go with raidz2, a metadata vdev should rather be a 3-way mirror. (Three 1 TB drives would be an option.)
Consider whether you really need a metadata vdev, or whether ARC (RAM) or a persistent L2ARC tuned for metadata (needs RAM first) would do (read workloads only).
Unfortunately, there’s no manual because documentation is always a work in progress.
Read it in parts, on a “need to know” basis: If you don’t understand a feature, you probably don’t need it.
That amount of RAM for not so much storage and what looks like the usual light apps is well beyond over “future proofing” and into “overkill” territory—but the only casualty is your wallet. Enjoy!
I would not use the 1 TiB NVMe for install, it can easily fit ont anything much smaller, (I use 16GiB Optane NVMe-s for the install) A 120GB sATA SSD costs like 20-30 USD now. Theoretically the speed is not much important, you can even boot up from a USB pendrive, however I would not do it. (you just have to save the configuration file of the system, and if it fails, you just do a clean install and import this save in like 5 minutes.)
Use the 1TiB drive for anything else, like apphome or VM home. (those need and actually benefit from the high bandwidth, that is available in an NVMe)
We don’t have a PDF version available, but each “book” in the documentation has a Download or Print link that takes you to a printable single-page html version of that documentation.
@DjP-iX I saw that before asking the question about a PDF manual. Unfortunately, when I clicked on the “Download or Print”, it took me to a blank page. It could be because of the Firefox Add-ons I use or my firewall/router’s ad-blocking. I don’t know though.
Here’s a couple of forum older postings in the resource sections that have relatively short pdfs (10-15 pages each). Although a bit dated, IMHO these are must reads, even before you start looking at the official documentation, which sometimes just tells you how to do things, not why you should do them (or more importantly, why not). The hardware guide in particular, is not necessarily useful for the specific models which are a bit old but for the logic used to evaluate what you are using and why.
Yeah the blank page should redirect you to the single-html build, so I suspect an extension is blocking the redirect. If you go to any of the landing pages for the books and append /printview/ to the URL, I think you should be able to get there directly.
I’ve basically learned everything from searching the forums.
