MP 7,1 Mac Pro 2019 NVME RAID drives failing to mount

[rhoulian]

Thanks - the 970 EVO Plus?

yep

 

[aventham]

yep

Thanks again - ordered!

 

[flowrider]

Hi guys,

My system is MacPro 2019
3,3 GHz Intel
AMD Radeon Pro W5500X 8 Go
128 Go 2933 MHz DDR4
Sonoma 14.4 (23E214)

Maybe the chances of getting this issue increase if you multiply the PCI cards ?

Nope _ I have two PCI cards (HighPoint SSD7103 & IOCrest IO-PCE2824-TM2) with five Samsung NVME SSDs and one Samsung AHCI SSD. I boot from one of the NVME Drives. I have never had a problem with drives not mounting👍🏻

Lou

 

[geruil]

Nope _ I have two PCI cards (HighPoint SSD7103 & IOCrest IO-PCE2824-TM2) with five Samsung NVME SSDs and one Samsung AHCI SSD. I boot from one of the NVME Drives. I have never had a problem with drives not mounting👍🏻

Lou

Hello,

These are hardware RAID cards? I am testing the Areca ARC-1686-2NOD card (hardware raid) and for the moment I have no problem (https://www.areca.com.tw/products/nvme-1686-nod.html).

On the other hand, I have problems with the OWC Accelsior 4M2 card (no raid hardware card). I am in discussion with OWC support.

It is also possible to open a bug report to Apple https://www.apple.com/feedback/mac-pro.html. Feedback Type: Bug Report.

Best regards,

Gervais

 

[aventham]

Hello,

These are hardware RAID cards? I am testing the Areca ARC-1686-2NOD card (hardware raid) and for the moment I have no problem (https://www.areca.com.tw/products/nvme-1686-nod.html).

On the other hand, I have problems with the OWC Accelsior 4M2 card (no raid hardware card). I am in discussion with OWC support.

It is also possible to open a bug report to Apple https://www.apple.com/feedback/mac-pro.html. Feedback Type: Bug Report.

Best regards,

Gervais

What drives are you running in your 4M2 @geruil?

 

[geruil]

What drives are you running in your 4M2 @geruil?

OWC Accelsior 4M2 : 2 x Samsung 990 Pro 2To (firmware update latest).
Areca ARC-1686-2NOD : 2 x Samsung 990 Pro 2To (firmware update latest).

Before it was : Crucial P3 2To but I suspect malfunctions (crash boot/kernel panic). So I changed.

 

[aventham]

OWC Accelsior 4M2 : 2 x Samsung 990 Pro 2To (firmware update latest).
Areca ARC-1686-2NOD : 2 x Samsung 990 Pro 2To (firmware update latest).

Before it was : Crucial P3 2To but I suspect malfunctions (crash boot/kernel panic). So I changed.

Are you experiencing the same symptoms with the Samsungs in the 4M2 as you were with the Crucials?
And do you know whether trimforce is enabled or disabled on your system?

 

[geruil]

Are you experiencing the same symptoms with the Samsungs in the 4M2 as you were with the Crucials?

The Crucial P3 2To were used in a Sonnet M.2 4x4 card :

• Sonnet M.2 4x4 card with Crucial P3 2To : when starting the Mac Pro no problem (never) but a few seconds later crash/reboot/kernel panic randomly. Using the bonnie++ command I was able to reproduce the problem randomly. I tested (Crucial software) the Crucial P3 2To and updated the firmware but still the problem.
• Sonnet M.2 4x4 card and OWC Accelsior 4M2 card with Samsung 990 Pro 2To : I only have the mounting problem at startup or restart Mac Pro. No crash, no reboot, no kernel panic.

 

[aventham]

Disabling trimforce seems to have stopped the mounting problem for me (Crucial P5s via 1M2 cards), but I still get kernel panics unless I unmount them before shutdown. Fingers crossed the 970 EVO Plus drives I'm intending to replace them with will put an end to that, as they are working well for rhoulian (albeit via 4M2 card and Softraid).

 

[rhoulian]

Yes, I confirm, still no issues for me, it's been a week now ! I shut down the MP every evening without unmounting anything.

 

[nopster]

Yes, I confirm, still no issues for me, it's been a week now ! I shut down the MP every evening without unmounting anything.

Me too!

 

[aventham]

Disabling trimforce seems to have stopped the mounting problem for me (Crucial P5s via 1M2 cards), but I still get kernel panics unless I unmount them before shutdown. Fingers crossed the 970 EVO Plus drives I'm intending to replace them with will put an end to that, as they are working well for rhoulian (albeit via 4M2 card and Softraid).

I just had another mounting failure. I'd still be interested to know if disabling trimforce helps anyone else, as it does seem to have made the problem less frequent for me; but it's evidently not the whole story.
Perhaps there is another storage-related maintenance process that can have this effect.
Has anyone tried installing NVMEFix?
GitHub - acidanthera/NVMeFix

 

[aventham]

Me too!

Glad to hear that @nopster! May I ask which drives you're using, raided or non?

 

[flowrider]

I'd still be interested to know if disabling trimforce helps anyone else, as it does seem to have made the problem less frequent for me; but it's evidently not the whole story.

I have never disabled TRIM👍🏻

Lou

 

[aventham]

I have never disabled TRIM👍🏻

Lou

Thanks @flowrider.
I read somewhere that it's supposed to be disabled by default (for non-Apple drives) but I'm not sure whether that's definitely the case as explicitly disabling it did seem to have a definite effect for me.
Which specific model of Samsung NVMe drive do you have?

 

[flowrider]

^^^^As posted above I have Six M.2 SSDs, one AHCI and five NMVE. I have Two Samsung 970 PRO, Three Samsung 970 EVO and one Samsung MZHPV AHCI.

Lou

 

[nopster]

Glad to hear that @nopster! May I ask which drives you're using, raided or non?

I'm using the Sonnet M.2 4x4 PCIe Card (Silent). As you can see in the screenshot, it is positioned in Slot 4, allocating x16, A option and Automatic allocation of bandwith switched on.

1. Disk used: Samsung SSD 2TB 980 Pro M2, Samsung 1TB same model
2. Disk format: RAID 0, done by Disk Utility, the disk is used for stuff storage, not to boot MacOS.
3. Options: FileVault off, other slots are empty on the machine

No problem at all after updating Sonoma. Hours switched on, switched off, put in sleep, awake... the drive remains active and functional.

 

[mattspace]

I'm using the Sonnet M.2 4x4 PCIe Card (Silent). As you can see in the screenshot, it is positioned in Slot 4, allocating x16, A option and Automatic allocation of bandwith switched on.

1. Disk used: Samsung SSD 2TB 980 Pro M2, Samsung 1TB same model
2. Disk format: RAID 0, done by Disk Utility, the disk is used for stuff storage, not to boot MacOS.
3. Options: FileVault off, other slots are empty on the machine

No problem at all after updating Sonoma. Hours switched on, switched off, put in sleep, awake... the drive remains active and functional.

AFAIK Slot 4 is fed by the unswitched PCI lanes that are allocated to the MPX slots, which might make a difference. It would be an interesting control to relocate the card to slot 5 and see if it remains as stable.

 

[nopster]

AFAIK Slot 4 is fed by the unswitched PCI lanes that are allocated to the MPX slots, which might make a difference. It would be an interesting control to relocate the card to slot 5 and see if it remains as stable.

Prior to updating Sonoma, the pci slot #5 is where the card gave me the problem of random and recursive hd disconnection. During tests, I moved the sonnet card on slot #4. Updating Sonoma now everything is ok, so I’m just a little bit scared about moving it!

 

[mattspace]

Prior to updating Sonoma, the pci slot #5 is where the card gave me the problem of random and recursive hd disconnection. During tests, I moved the sonnet card on slot #4. Updating Sonoma now everything is ok, so I’m just a little bit scared about moving it!

If the problem is related to the PCI Switch, it would be a good diagnostic test.

 

[aventham]

So far so good with my Samsung 970 Evo Pluses in place of the Crucial P5 Pluses (which also seem happy enough relegated to USB-C caddies).

 

[aventham]

AFAIK Slot 4 is fed by the unswitched PCI lanes that are allocated to the MPX slots, which might make a difference. It would be an interesting control to relocate the card to slot 5 and see if it remains as stable.

Apologies if I'm talking at cross purposes, but @nopster's screenshot indicates that slot 4 is switchable between pool A and pool B when it isn't being dedicated to MPX.

 

[mattspace]

Apologies if I'm talking at cross purposes, but @nopster's screenshot indicates that slot 4 is switchable between pool A and pool B when it isn't being dedicated to MPX.

You're right, my mistake. I had it in mind that all the slots within the two MPX bays were from the unswitched source, but it's only slots 1 & 3 that aren't switched.

 

[geruil]

Hello,

Last Friday (March 15), i downgraded from macOS Sonoma 14.3.1 to macOS Ventura 13.6.5 (fresh install). For now, I have no problem.

Best regards,

Gervais

 

[aventham]

So far so good with my Samsung 970 Evo Pluses in place of the Crucial P5 Pluses (which also seem happy enough relegated to USB-C caddies).

Ugh - mounting failure on day 4. The one drive that showed up was the Crucial P5 I had plugged in via a caddy!
And now I seem to have to shut down and restart, rather than just restart, to get the PCIE Samsungs back!
But if I don't get any more kernel panics it's still a pretty decent net gain I guess.

Perhaps some sort of maintenance routine occasionally tries to run during startup, but at an inopportune moment that clashes with the PCIE drives being mounted? There used to be these maintenance scripts, but I'm not sure whether there still are - the command given in the article to look for them doesn't turn anything up for me:
How to carry out MacBook maintenance: all you need to know (macpaw.com)
I've tried running the following command manually -

sudo periodic daily weekly monthly

- but it completes instantly and doesn't generate any such logs.