P44 Pro nvme controller is down will reset

[andy]

NVMe (P44 Pro, model SSDPFKKW020X7) sometimes just disconnects and stops working, smart(attached) reports everything being normal, temperatures are also within reason (around 45-50C, graph of a few minutes before the incident attached). Drive disconnects after about a week or two of uptime (although it was fine for the first month and a half of use). OS is arch linux running kernel version 6.4 (and a few older ones, but this is for the most recent occurrence).

Firmware is on the latest version (checked with the update tool).

I removed the serial number from my smart output (just to be safe), but can send it if needed.

I saw someone with what looks like a related issue (on windows, I assume it's blue-screen-ing due to a spontaneous disconnect): https://community.solidigm.com/t5/solid-state-drives-nand/p44pro-too-hot-to-lose-disk/td-p/24074

Tempurture of nvme drivekernel logs of drive disconnectingsmart output of drive after rebooting

[oscarfowler]

Just about 3 months since I replaced my P44 with a Samsung 990. For anyone wondering: I've not had a single reboot or crash since replacing the drive. And this is despite the fact that I've gone through some very heavy usage periods in that time.

[SolidigmGleb]

Dear oscarfowler,

Thank you for the information provided. We are sorry that you have a problem with our drive. 

We would like to suggest you use a standard driver, as it might help to solve your issue. Anyway, you always can  contact us to create a case by using this link- Create Case  · Customer Self-Service (solidigm.com)  

Kind regards,
Gleb
Solidigm Customer Support

[SolidigmGleb]

Dear oscarfowler,

Thank you for your insights regarding the P44 Pro drives. It's intriguing that different usage patterns and motherboard connections could influence the drive's behavior. Switching the NVMe socket from one connected directly to the CPU to one connected to the chipset, as you did, is a worthwhile experiment to see if it resolves the issue. 

Kind regards,
Gleb
Solidigm Customer Support

[andy]

> Was there ever a resolution to this?
TL;DR: Same thing that happened to you, I switched to samsung and haven't looked back (no issues in the last 8 months)

Full story:
A ticket was opened and I talked to support. They had me send a smart log using one of their tools, but no errors were found there. They had me try a few things like re-installing the OS, switching the port the NVMe was plugged into, try it in a different computer - none of which really worked (but was hard to test because it took weeks for the issue to manifest).

Eventually they sent me an email stating that I have an unsupported configuration and a link to the warranty - having read through that I noticed that it said that they do not support "any computer ... *that supports* workloads ... of more than one concurrent user or one or more remote client device concurrently". Which to me sounds like they don't support WIndows 10/11 Pro (you can RDP onto a machine - remote user; you can share files on the drive over samba). So it "supports" multiple users or concurrent client devices accessing the drives. The "supports" bit means that you don't even have to use those features for them to tell you that you're out of the warranty.

I did end up giving the drive away to a friend (having told him this story) in hopes that the windows drivers would function better... but I guess not.

-----

Having written this again, I would be interested in trying to set the NVMe drive to use a PCIe4 x1 connection (instead of x4 lanes like usual). This should get rid of the concurrent accesses (which seems to be a know bug given the wording of the warranty). But that would negate the performance benefits of using NVMe.

[oscarfowler]

That exception in the warranty is crazy, since Linux, Windows and Mac all support concurrent users in one way or another (e.g. opening up SSH).

Thanks for posting your reply - I wasn't willing to waste the time re-installing the operating system, testing with a different drive, or putting the drive in another machine. All of those tests are an onerous burden on the customer, especially considering you'd have to wait weeks or longer after the change to confirm whether any change had occurred. If I was going to spend the time swapping out the drive and/or reinstalling the OS, I was going to do it with a different brand drive and greatly improve my chances of not having to do it again.

I saw your post before I called their support. Though it's possible that it's a software bug that affects both Linux and Windows, I think it's more likely a hardware bug. I guess we'll never know since they seem disinterested in diagnosing the issue.