Dang I sent an Offer for 6 without realizing that
.Hopefully I lowballed low enough (80 USD / Piece).
Enterprise SSD "small deals"
- Thread starter ca3y6
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
.Hopefully I lowballed low enough (80 USD / Piece).
QLC isn’t necessarily bad. Most people will never go near the limits of enterprise TLC endurance.
Looking at the data sheet for the Micron 5210 ION they’ve pretty lousy write IOPS and write speed. Definitely a WORM drive. Endurance is a bit lower than the 5100 ECO.
I guess they were a pretty good deal new if you upgraded from HDDs and had mostly WORM data. Still, even this post might be a decent deal depending on use case.
I guess they were a pretty good deal new if you upgraded from HDDs and had mostly WORM data. Still, even this post might be a decent deal depending on use case.
Quite decent Micron 5210 ION 3.8 TB Specs but these are more read intensive oriented as well. Would be great for a Plex/Jellyfin usage.
For the moment I am only selling a couple of 10DWPD 3.2TB drives:Where are you selling your 3.84TBs
Western Digital SS300 3.2TB SAS SSD 10DWPD Write Intensive 100% Health | eBay UK
Find many great new & used options and get the best deals for Western Digital SS300 3.2TB SAS SSD 10DWPD Write Intensive 100% Health at the best online prices at eBay UK! Free delivery for many products!
www.ebay.co.uk
They are good deals but not great deals!! I have a handful of 3.2TB U.2 and a big pile of 3.84TB SATA that I keep around for a hypothetical future project.
Only one each?For the moment I am only selling a couple of 10DWPD 3.2TB drives:
Western Digital SS300 3.2TB SAS SSD 10DWPD Write Intensive 100% Health | eBay UK
Find many great new & used options and get the best deals for Western Digital SS300 3.2TB SAS SSD 10DWPD Write Intensive 100% Health at the best online prices at eBay UK! Free delivery for many products!www.ebay.co.uk
They are good deals but not great deals!! I have a handful of 3.2TB U.2 and a big pile of 3.84TB SATA that I keep around for a hypothetical future project.
My very limited understanding is that 1 bad sector isn't a big deal. The price is concerning though. Makes me think the seller knows something they aren't letting on, either that or it's a great deal right?
Usually once the first bad sector appears more are on the way. These enterprise drives should have enough buffer cells to reallocate so that means it's burned through all of it? (Don't quote me on this)
Oh I see, so having a bad sector means it's already had enough reallocated sectors to burn through the extras. Hard pass then.
Not necessarily, if the sector fails to read the drive has to report it so that raid controllers or software can do the right thing, but then a future write would go to a fresh page of NAND from the reserve and work fine. It's sort of impossible to tell exactly what the seller means without the full stats from a drive.
And I guess we could do our own Overprovisioning, although to be honest since I never did it myself via
hdparm, I don't know if there is a difference between:- Doing via
hdparmand directly modify the Capacity that the SSD reports to the OS - Doing via Partitioning (e.g.
partedorgdisk) and create a Partition smaller than the entire Drive
Typically I always do Partitioning and always leave a small Fraction of the Drive (e.g. 512MB for a 0.5TB - 4TB Drive) unpartitioned at the End of the Drive, to account for slighly different Number of Sectors between Drives. I cannot say I ever saw the Need for it though, they seem to perfectly match in Terms of Sectors and if there are Differences, it's typically 3.84TB vs 3.6TB or 1.92TB vs 1.6TB or something like that, something that a meager 512MB cannot compensate at all.
I'd wonder for Overprovisioning if, using a Partition "Type" of Overprovisioning (e.g. leave 20% of the Drive unpartitioned), that would achieve the same Result as doing it via
hdparm.Since I don't really look at smartctl all that much (apart from TBW and Temperatures), but rather rely on ZFS Checksum Reports, probably the end Result, at least for me it's the same. Basically when the OS / ZFS thinks the Drive is full, there are actually still quite a few Sectors available, which could in Theory be used for Wear Levelling. BUT I guess that since the Drive itself does NOT know that we want that Part of the Drive to be used for Wear Levelling, it most likely would still generate SMART Errors.
Any Thought
?As far as I know SSD controllers don't parse the partition table at all, so unless you're using NVMe namespaces to over-provision you just want some region that won't ever take writes from the host. I don't think that really gains you much on an enterprise drive unless you're trying to get better sustained random write performance out of a 'read intensive' type drive.
Alright. I'm not tempted to touch a SSD which already has at least some Sectors that are Bad
.I just had a HDD fail on my Server Today without ANY Warning at all (which is weird in my Experience)
: just disappeared from the OS, no Click, no ATA Errors in dmesg, nothing. Just some ZFS ZIO Errors which is normal since the Drive couldn't be written to (and zfs failed it from the affected VDEV).