Hello guys,

 

so I replaced my old SATA SSD Cache Drive with a shiny new NVMe M.2 SSD. 

Unraids drive log tells me this:

 

-Power on hours 41 (1d, 17h)

-Unsafe shutdowns 64

-Media and data integrity errors 0

-Error information log entries 2,016

 

Should I be worried about the error information logs? count is still rising. I searched the internet for this message and some say it's totally normal and ignorable.

 

thanks for your help

Best have NVMe full SMART report, then compare to other which have same NVMe.

So now I had this in my drive log:

May 28 11:33:25 unRAID emhttpd: shcmd (471): mkfs.xfs -m crc=1,finobt=1 -f /dev/mapper/nvme0n1p1
May 28 11:33:25 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 4160
May 28 11:33:25 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 8392767
May 28 11:33:25 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 16781374
May 28 11:33:25 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 25169981
May 28 11:33:26 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 33558588
May 28 11:33:26 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 41947195
May 28 11:33:26 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 50335802
May 28 11:33:26 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 58724409
May 28 11:33:26 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 67113016
May 28 11:33:26 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 75501623
May 28 11:33:26 unRAID root: meta-data=/dev/mapper/nvme0n1p1 isize=512 agcount=4, agsize=58605652 blks
May 28 11:33:26 unRAID emhttpd: shcmd (473): mount -t xfs -o noatime,nodiratime /dev/mapper/nvme0n1p1 /mnt/cache

 

and I will also add my full SMART report here.

 

The device is a Corsair Force MP510 960GB 

Force_MP510_1.txt

Those shoud be abnormal, Uraid and NVMe both have error.

I haven't Phison controller SSD for ref. and try search on internet also not found long run-time SMART info. ( ID : 0F ) 

 

Had this error again last night. My SSD-Trim chron runs at 2AM so it was probably caused by that.

 

so should I replace my device or is this still okay? Ca anyone confirm?

 

May 28 21:50:12 unRAID emhttpd: shcmd (471): mount -t xfs -o noatime,nodiratime /dev/mapper/nvme0n1p1 /mnt/cache
May 29 02:00:01 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 266962616
May 29 02:00:01 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 275351223
May 29 02:00:01 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 283739830
May 29 02:00:01 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 292128437
May 29 02:00:01 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 300517044
May 29 02:00:01 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 308905651
May 29 02:00:01 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 317294258
May 29 02:00:01 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 325682865
May 29 02:00:01 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 334071472
May 29 02:00:01 unRAID kernel: print_req_error: I/O error, dev nvme0n1, sector 342460079

So I bought a second SSD. same brand same model same size and the -Error information log entries were gone for days.

Then my ssd trim cron kicked in and boom, same "issue". Getting the same I/O errors and my -Error information log entries started to rise.

So I am still not sure if this is ignorable. My system runs fine though. 

Maybe worth mentioning is, that my server does not have a dedicated M.2 slot, but I used an adapter to PCIe. Tried two models, no difference.

 

3 hours ago, DelSol said:

but I used an adapter to PCIe.

Won't relate raiser adapter. Likely relate driver issue which have problem when execute TRIM.

 

Nowadays, seems different NVMe have problem report

 

SMI, passthrough issue

Marvell, don't remember

Phsion, TRIM issue ?

Samsung, any ？

 

Edited June 3, 20197 yr by Benson

16 minutes ago, Benson said:

Samsung, any ？

Samsung are known to work fine, at least AFAIK.

Or Intel.  Their SSD's are good for this case.

 

IIRC some Intel NVMe devices also have trim issues in some cases.

Intel current main stream controller was SMI.

It seems like in most other cases TRIM does not work at all, for me it's only some sectors per run failing...

So should I switch to another SSD? 

Samsung 1TBs are pretty expensive, the intel 660p is about the same price as the corsair, but slower in I/O and other benchmarks. 

Is it really worth switching?

Intel 660p is QLC, that's why it's so cheap.

So for everyone this is (probably) the solution to my issues:
First of all my ssd is encrypted via LUKS, I did not know that this makes any difference. 

LUKS itself disables any TRIM commands for security reasons.

I don't know why I'm getting hundreds of error messages for every block TRIM tries to modify while others just get one error message like "trim not supported" or "device not accessable".

Also I never had any issues with my encrypted sata ssd running trim...?

anyway, I'll stick with my corsair ssd for now, disabled the TRIM cron and will watch its performance the next weeks.

 

Are you running latest Unraid? It has support for trim on encrypted devices.

I'm on latest stable. 6.7.0 is this something new in the beta?
If not, then it's probably caused by my old PCIe controller...

It was added starting on the 6.7 rcs.

I was considering this SSD for my RAID 1 cache drives. Is this only an issue when you want to use encryption on the SSD?