Need recommendation for External HDD

I know, that external 2,5" USB HDD - worst storage for 24/7 running node, but at the moment on one node i only have this option to go.

I would like to buy something like WD RED 2,5" 4TB and put it in to proper enclosure, but they have only 1TB as 2,5". But for the WD Passport 2.5" they have 5TB drive! As well many other manufactureres do have 2.5" 4-5TB drives. Of course they are not dedicated for 24/7, but at least with 3 years warranty.

Any recommenations for such a external drive? Maybe somebody have good or bad experience with specific disks ?

TY

2.5" is limited in capacity for obvious reasons. This 5TB drive seems to be an SMR drive, stay away from those if possible. If you need USB interface I suggest you get a good 3.5" drive and attach the interface converter to that.
Regarding experience I have had bad experiences with Seagates, everything else worked well for me personally.

And may I ask, why is USB the only option for you? USB actually may not be bad in itself, it’s just that the drives that come in USB enclosures are usually the lowest quality from that manufacturer.

IMHO All external usb HDD (mostly 2.5’’) above 3Tb - SMR.
So, no matter what kind of external usb HDD to buy.
Disable write cache and move database to another, non storj usable, disk.

https://www.anandtech.com/show/14849/wd-my-passport-5tb-das-review-a-compact-capacity-play-sans-smr-hassles

Please, read this article:

1 Like

Thank you. I’ve read a lot about SMR problems. Looks like i will change the plan. Will upgrade better another NODE with external CMR HDD.

@BrightSilence will move our conversation here.

Do you use them in RAID/NAS? If no, why did not went for Seagate SkyHawk? They a bit cheaper, but also designed for 24/7. I believe one NODE will not make such a huge load on drive in compare to the business with ±50 users using drive it in the NAS.

Yes, I use them in an SHR2 array. I bought the Exos because they were actually cheaper than Ironwolf and for some reason Synology lists the Ironwolf Pro 16TB as not compatible. Skyhawk is barely cheaper and not optimal for use in an array, so I didn’t consider it.

For array not optimal i agree.
But for usb enclosure & RPi probably will be the best option.

A. For this particular node the only option, because it’s RPi.
B. Enclosures it self also can be buggy even with the good drive.

Then get a good interface adapter, there shouldn’t be anything inherently bad with that, and there is no way to exceed the bandwidth specs anyway.
The adapter dying is much better than a drive dying, they rarely take out the data, if the drive dies (completely) it takes out the data by definition. In any case, get the drive and the adapter separate from each other preferably.

I’m on the way of receive a WD_Black p10 5Tb external drive. It is supposed to be an “intensive use” drive - sold as a gamer drive for consoles and PCs, but good price per Gb in an AWS offer a few days ago.

As far as I have read, a 2.5" 5Tb disk is SMR for sure, but I cannot find it listed anywhere as SMR. Can this be checked anyway when I receive the unit before it´s too late?

Also, is SMR a problem for a disk attached to RPi4, as it would not potentially receive as much data as a faster node?

WD Black is not rated for 24/7, but probably it is the best drive from all not rated 24/7.

I haven’t seen yet a big difference betweed i7, RAID6, LSI, 16GB RAM vs RPi4 with external drive. There might be a slighly difference, but it is very small. Maybe we should wait for a ultra havy load in the network to see differences better.

While upgrading one of the nodes, i’ve put for a few days 3 nodes on a single RPi4 with total 12TB of data. CPU load was 20% at the max. Memory used 400Mb (out of 4Gb). I only seen some problems with IOWAIT (for up to 10min), but only when all 3 nodes are starting up at the same time.

1 Like

Yes. The IOWait goes up to 80% and the entire system is braked. Especially the access time to the hard drive, which means you lose the race for the piece. I could see that again and again that the Docker process ran between 80- and 120% CPU utilization. Since I replaced the SMR hard disk, the problem has never occurred again.