I was hoping to get some feedback here on a new set of drives for my pool. I currently have 6 4TB WD Red’s and would like to upgrade to the highest capacity I can. I read in the community hardware guide 2a document though that it’s not recommended to go over 5400 RPM drive for reliability and temperature performance. From what I’m seeing that limits me to a 6-8TB drive. Anything over that seems to always be 7200 RPM.
What’s everyone’s thoughts on this? Should I just stick those smaller size capacities? It’s currently in a mini-itx case so don’t want to get myself into an overheating issue.
Not specifically but isn’t that a 7200 RPM drive? Has the TrueNas community gone away from discouraging 7200 RPM drives? The hardware guide 2a is a few years old so maybe it’s outdated by this point.
I, for one, wouldn’t go for it. Usually, bigger drives are cheaper (per TB) nowadays. If you face some overheating issues, you can always just drop some fans (depending on the case this may look DIYish, though).
Then that document is severely outdated because there are no longer any (new) NAS-grade 5400 rpm drives. But if you’re interested, WD can sell you “5400 rpm-class” drives which spin at 7200 rpm; they may even have a bridge to go with it…
Go with 7200 rpm helium-filled drives. Or with SSDs if capacity is low enough.
@etorix beet me, I was going to say that the reference is outdated and it is very difficult to find a 5200-5400 RPM drive these days. This means you just need to purchase based on reliability (or perceived reliability) and the manufacturer specifications. Of course the cost plays a huge part in our selections of drives as a home user.
There is a huge difference between 4TB and 24TB drives from several perspectives. I say 24TB because I believe (I could be wrong) but the 26TB drives are SMR, however the ST24000DM001 (24TB) is a CMR drive. There are others as well, but you MUST know what you are purchasing. Warranty period is also a big factor at these prices.
One major concern is Resilver times. Of course we are assuming that if you have more storage capacity, that you will add a lot more data to it. Scrubs will take much longer, and a SMART Long/Extended test will take a very long time.
You did not specify your pool layout, hopefully it is a RAIDZ2 at a minimum. Drive failures during resilvering happen more than you would think, which is why, unless the old drive has completely failed, you would add a new drive, resilver, then remove the old drive. This is the safe way to do things.
If cooling is a real issue, you could consider a case modification if you are handy with some tools, and patient. If you were thinking about this approach, toss a few photos of the system, provide the case make/model, and I could provide you good solid recommendations on how to modify the case to provide improved airflow/cooling to the drives and possibly other components. A mini-itx case is fairly small. All of my modifications are nice looking and look like it could have been manufactured that way, assuming you have to cut a hole. Sometimes only adding a piece of cardboard or piece of foam tubbing is needed to redirect airflow. And you can always make it temporarily out of materials and replace it with a nice 3D printed part.
I can see why a guide might recommend 5400 RPM drives, but that guide’s recommendation shouldn’t be confused as a rule etched in stone that you are required to follow.
I don’t think believe that the actual TrueNAS Community would discourage the use of 7200 RPM drives. In fact, I bet if we had some telemetry data from all the TrueNAS installations out there we’d find that hard drives over 5400 RPM massively outnumber 5400 RPM drives.
While they have done a great job with this research, it is probably not applicable to home use. They have their drives temperatures between 15°C and 30°C. My drives run for 36-42°C (mostly dependent on ambient temperature).
This reminds me of similar research made by Microsoft (IIRC). And they didn’t find significant difference in AFR between drives with 18°C and 28°C.
Same thought is in the first comment for this very research:
Jack McDonald9 years ago
You can’t measure between 21 and 31 degrees celcius and say there’s no correlation. That’s like saying cigarettes have no correlation with cancer because 0 a month and 1 a month have the same results.
You need to test right up to the ‘Operating Temperatures’ of ~60 degrees.
So, unless you have a 24/7/365 air-conditioned server room, this research is not applicable.
If there was a causal effect between heat and drive failures, it’d be present in a large enough sample size. It might be negligible at the lower temperatures in the study–but it’d still be present if it were statistically significant.
It’s actually applicable, but how applicable is probably requires a degree of independent. When you consider the published operating temperature ranges for hard drives, I’d argue that your home is probably much more like a “24/7/365 air-conditioned server room” than you realize it is.
In order for it to be “not applicable”, there would need to be a similarly-sized study proving otherwise. Unfortunately, that probably won’t be happening any time soon.
The conclusion of that study would actually establish a causal correlation between survival and the height of the fall, wouldn’t it?
Would you feel comfortable setting your refrigerator to 45 degrees F because there was not much of a difference of spoilage when 33 to 35 degrees was tested? Once you pass a certain threshold, the rate of spoilage jumps up.
It’s not that Backblaze’s results don’t establish a correlation of drive temps and drive failures, but those who run their HDDs under higher ambient temperatures can’t really gauge how much of a difference it applies to them.
In fact, some of Backblaze’s data shows that increasing the temps decreases the failure rate. Look at the trend from 24 → 30 degrees.
“What we can deduce is a correlation between drive temps and failure rates. As the drive temps increase, the failure rates decrease!”
I agree with @swc-phil and the highest rated comment on their blog: This data has no use for most home users, whose drive temps easily reach anywhere from 34 to 40+ C.
Why is that so? I do believe that at 100°C, the drive wouldn’t survive very long. And yet there is no sign of it in the aforementioned statistics.
Nonsense.
“According to the internet poll, we have 100% population internet coverage” – no counter-study needed to say that this internet-poll is not applicable for finding the internet coverage percentage.
First of all, it’s worth pointing out that you’re moving and narrowing the goal posts in your example:
Backblaze found no correlation between drive temperature and failure rate. Your example has a correlation between temperature and spoilage.
Backblaze’s study covered an much wider temperature range than the 3-degrees of Fahrenheit.
But I wouldn’t change my fictional refrigerator’s temperature because as you clearly stated: they found a causal correlation between spoilage and temperature rising from 33 to 35 degrees. At the very least, I’d know that raising my refrigerator to 45 would likely be as bad–if not worse.
I can’t agree with this more–except that the prevailing sentiment here in this thread that because the difference can’t be gauged for some reason the opposite of the study must be assumed as fact.
You’re manipulating and misrepresenting that image to try and prove a point for an imaginary “what-if” scenario that isn’t happening
This entire thread seems to have veered into the “no use” realm.
@chrisolson91, my advice is unchanged. The additional potential heat generated by a 7200 RPM (compared to a 5400 RPM drive) is irrelevant and you’re absolutely fine to use them in your NAS.
I could have left 31C in there. Why didn’t they shown us data for drives up to 55C? How is it other than the often cited Backblaze data that the mantra has otherwise been “improve airflow and cooling for your drives”? Was it always just an old wives tale?
Which raises the question, what use is it, as “large” as the number of drives being studied, when they are only collecting data for a range of temps that home users are unlikely to operate.
We might learn that after a certain point, not only do higher drive temps correlate to failure rate, but that the effect increases with every iteration above a certain threshold.
There’s a reason why the top rated comment on their blog states the same sentiment.
