Samsung Evo 840 Fixes
Queued TRIM seems to be a problem with this kind of drive (with the latest firmware), and a “recent” kernel (4.1.x) seems to have this fixed this without disabling NCQ completely. Is that patch backported to the
“mainline” CentOS 6 or CentOS7 kernel?
7 thoughts on - Samsung Evo 840 Fixes
Can you provide any references to the specific bug you’re referring to?
This is the first I found. http://linux.slashdot.org/story/15/07/30/1814200/samsung-finds-fixes-bug-in-linux-trim-code
and https://bugs.launchpad.net/ubuntu/+source/fstrim/+bug/1449005
We have used that drive for 2 years with no issues.
Amazing how much trouble people can get into when they tweak stuff…
-Joe
—–Original Message—
That bug isn’t specific to Samsung drives. It affected all SSDs used in raid0 and raid10 configurations.
That’s a different bug, entirely. The bug above was a Linux bug, where this is a drive firmware bug. It’s worked around in Linux by blacklisting NCQ TRIM for Samsung’s drives.
The ATA_HORKAGE_NO_NCQ_TRIM flag was introduced in 3.13. I don’t see it in CentOS 7’s kernel, so it’s almost certainly not in C6 either.
You’re probably better off disabling NCQ or just testing to see if you need TRIM at all. Particularly if you leave some space unallocated on the drive (TRIM it if you were previously using it), many newer drives manage free blocks very well, and see little or no performance hit for systems that don’t use TRIM.
Huh, like upgrading the firmware?
Yeah, I have about 50 of them in raid cards and doing software raid.
Never done that. :)
No need.
We have yet to have one fail in 2 years :)
—–Original Message—
Am 11.11.2015 um 12:36 schrieb Joseph L. Brunner:
so, the next dozens will come shortly :-)