The Linux Thread - The Autist's OS of Choice

  • 🐕 I am attempting to get the site runnning as fast as possible. If you are experiencing slow page load times, please report it.
♂CANAM productions♂ said:
BTFS/JFS, same issue is present on NTFS and XFS. Seems to be a NVME controller bug as it is showing the controller is crashing (Removing after probe failure status: -19). See this kernel bug (Unresolved since 2017). I've got a Samsung controller which is known to cause issues.
Click to expand...
Did you check for known incompatibilities between your motherboard and your nvme drive?

If it's one that connects directly to your mobo you also might want to play around with reseating the thing and giving the screw a little more slack. I had one that kept dying because a very tiny amount of thermal expansion would yank it out of the port.
 
HootersMcBoobies said:
Last SSD I had that failed was a regular SATA, and that was ~10 years ago .. no errors in the logs though.

I have had a ton of ZFS errors once and it ended up not being the drive at all, but the system memory. You run memtest86?
Click to expand...
Drives either fail early on, or late. Failure in “middle age” is relatively rare.
IMG_0092.png
Very few consumers ever use a drive hard enough to actually wear it out. That goes for both flash and mechanical storage.
Memory errors often show up as corrupted files on the drive. ZFS reacts to these as if they are corruptions on the drive itself, and scrubbing the drive at this point basically means overwriting random bits of them with nonsense.

It’s why I’m adamant that ECC needs to be a standard feature. Just because we’re consumers doesn’t mean we deserve to be stuck with intentionally faulty hardware.
 
  • Like
  • Agree
Reactions: Instant_Pot_User, Vesperus, Fabio Socialist and 7 others
Susanna said:
ZFS reacts to these as if they are corruptions on the drive itself, and scrubbing the drive at this point basically means overwriting random bits of them with nonsense.
Click to expand...
Chances are it will rebuild from stored parity bits just fine but it could end in corruption.

The wild part is that DDR5 has on-chip ECC but that isn't exposed to the bus.
 
Mikoyan said:
Chances are it will rebuild from stored parity bits just fine but it could end in corruption.
Click to expand...
The point of redundancy and checksumming is that you want the data to stay intact. If your memory is bad, using ZFS is actively worse than just LVM RAID and ext4, because that at least won't actively corrupt itself by trying to repair imaginary errors, it'll just crash.
Mikoyan said:
The wild part is that DDR5 has on-chip ECC but that isn't exposed to the bus.
Click to expand...
DDR5s ECC is all about cost savings. On-chip ECC means you can lower the bar for quality control, but connecting it to the bus would raise the cost of production. The difference is utterly marginal, though, so I consider it a spit in the face. "You want ECC? Fine. We won't let you use it though."
 
  • Agree
Reactions: Fabio Socialist and Owlbear
Susanna said:
Very few consumers ever use a drive hard enough to actually wear it out. That goes for both flash and mechanical storage.
Click to expand...
Yeah, it's unlikely unless you run across a bug where either a program or the kernel just keeps writing and overwriting the same sectors over and over 24/7.

This actually kills your shit very quickly.
 
  • Agree
Reactions: Chen Stirner and Susanna
std::string said:
Did you check for known incompatibilities between your motherboard and your nvme drive?

If it's one that connects directly to your mobo you also might want to play around with reseating the thing and giving the screw a little more slack. I had one that kept dying because a very tiny amount of thermal expansion would yank it out of the port.
Click to expand...
Yes. Tried both. Seems to be said firmware issue with linux. It's fine now after passing that command.
Edit: I spoke too soon. Running into the issue still. Seems like controller issue.
 
Last edited:
  • Informative
Reactions: std::string
I was frustrated that there is no straight answer to the question of how to install some programs on a separate drive. The only good solution I see is the usage of SymLinks.
 
Precision Anal Strike said:
I was frustrated that there is no straight answer to the question of how to install some programs on a separate drive. The only good solution I see is the usage of SymLinks.
Click to expand...
Out of curiosity, what do you believe is a benefit to installing some programs to another drive, instead of just moving /home and using stuff like appimages?
 
  • Like
Reactions: Marvin
Soo... That drive issue I'm dealing with, if any runs into the same issues.
There's several separate issues that are present in true bug form.

1) Failure of the controller to enter power saving mode (The fix I posted earlier). Supposedly fixed after 4.49 but some still have that issue.
2) Suspend support failure on AMD based CPUs which throws a "nvmeX: Removing after probe failure status: -19" . Setting Iommu kernel parameter as "soft" or "fullflush" may fix this for some reason I can't figure out why.
3) Bad blocks on my NVME device the controller is writing to (it shouldn't) because of the way I partitioned BTFS, NVME drive block size is not equal to the BTRFS block group size. Phisons controller is absolute garbage.
Hoped this helped anyone who runs into similar issues.
 
  • Like
  • Informative
Reactions: SphinxItOver, Another Char Clone, std::string and 3 others
Precision Anal Strike said:
I was frustrated that there is no straight answer to the question of how to install some programs on a separate drive. The only good solution I see is the usage of SymLinks.
Click to expand...
♂CANAM productions♂ said:
Soo... That drive issue I'm dealing with, if any runs into the same issues.
There's several separate issues that are present in true bug form.

1) Failure of the controller to enter power saving mode (The fix I posted earlier). Supposedly fixed after 4.49 but some still have that issue.
2) Suspend support failure on AMD based CPUs which throws a "nvmeX: Removing after probe failure status: -19" . Setting Iommu kernel parameter as "soft" or "fullflush" may fix this for some reason I can't figure out why.
3) Bad blocks on my NVME device the controller is writing to (it shouldn't) because of the way I partitioned BTFS, NVME drive block size is not equal to the BTRFS block group size. Phisons controller is absolute garbage.
Hoped this helped anyone who runs into similar issues.
Click to expand...
That's a shame, I was considering putting money to get a Phison drive for bootleg VRAM.
 
Lone Wandering Courier said:
Yeah I've been using Linux Mint for a while now. It's a decent distro and a new major version is coming soon.
Click to expand...
there's not a lot different between 22 and 21.3 aside from improvements to the underpinnings and an upgrade to the latest version of Ubuntu, but lack of shiny new eye candy aside it being a stable upgrade seems good.
 
someonenew said:
dose anyone here use mint os?
Click to expand...

I am currently using LMDE (Linux Mint Debian Edition) but I might move over to mainline Cinnamon or EDGE, I'm curious as to how ZFS works with Mints own configuration. Otherwise, LMDE is pretty solid despite being the oddball.
 
Last edited:
You must log in or register to reply here.
Back