StarTech PEXM2SAT3422 initial problems

Background

The card:

The StarTech PEXM2SAT3422 is a PCIe 4x card featuring 4 SATA ports as:
2 x SATA connector
2 x M.2 NGFF connector

It uses the Marvell 88SE9230 chipset

The card features RAID 1, 0, 10, as well as SATA pass-through.

Upon initially deploying the card, I experienced some very serious problems with it

My setup:

Server is a SuperMicro X9SCM in a 2U half-depth chassis.
3Ware Card exists with 2 1TB WD Black HDs.
Motherboard Intel SATA has 2 Kingston SSDs installed in a RAID1 (Microsoft Dynamic Disk)
OS is Server 2012 R2

This server has been rock solid for 3 years.

About 1 year ago:

I added an Ablecomm SATA card that has 2 M.2 connectors. It has housed a single Crucial CX300 M.2 SATA SSD. This setup has been rock solid with no hiccups whatsoever.

Today:

I replaced the Ablecomm PCIe card with the StarTech card and moved the CX300. I also installed a 2nd 1TB CX300 alongside. The plan was to mirror the 2 CX300s and start using them for production workload.


Here are the problems I encountered:


  1. One M.2 disk disappeared (showed offline in the Marvell tray tool). This degraded the RAID1. I had to shutdown the server & power back up to see the disk again.
  2. Cannot update the firmware of the Crucial MX300 1TB SSD while connected through this PCIe card. This is true whether in RAID mode or pass-through. (this seems like a glitch, as I would really expect pass-through mode to expose the drive directly off of the Marvell controller to the OS just like any other storage controller).
  3. Both CX300s disappeared while the RAID1 was rebuilding after my reboot (after the 1 drive had initially disappeared).
  4. The Marvell management utility Windows service stopped running (I believe it crashed). I had to restart it manually to discover that the drives were missing entirely.
  5. Upon rebooting the server, we get a black screen with txt cursor and the OS never loads.

Theories & Additional Details:

Problems 1-3 happened while the disks were under max load transferring VHD files AND I was using the Maevell management utility (web interface) to view various data (health & configuration information). I don't know whether my viewing of this info while the controller is under load should cause a disk to drop off. But it is likely that the disks would have fallen offline regardless of whether I was viewing info in the management interface.

The 2 SSDs' firmware are not on the same version. They are on 1 revision difference.

I suspect the current black screen at boot up is actually the controller rebuilding the RAID. This is because the LEDs on the card for both SSDs are flashing like crazy. If this is true, it means the card halts boot-up (potentially for hours) for array rebuild.

Game Plan:

Let the rebuild complete (assuming that's even happening at this point).

If it completes:

  • Try to figure out how to update the firmware on the 2nd Crucial MX300
  • Call StarTech & find out whether I'm doing anything wrong.
    • Card firmware updates, drivers (though I already installed their driver), etc.

If it doesn't complete:

  • Power off
  • Pull out the card
  • Remove the 2nd (older firmware) SSD
  • Reinstall the card
  • Move the VHD files that are on it back to other storage.
  • Contact StarTech, try firmware updates, etc.
    • (Leaving it in this state should at least verify the card stability for a single disk.)
  • Move the Ablecomm card & old SSD to a different PC & update firmware there.
  • Verify whether I have the correct version of the Crucial utility for firmware update (I updated the other MX300 SSD using this same card 1-2 weeks ago in a Windows 7 PC).

Final Thoughts & Conclusions:

It seems I should have gone with a real RAID controller. I realize technically this is a real RAID controller. But I mean something like an LSI or SuperMicro that is rock-solid.

This StarTech card has raving reviews on Amazon and it is really surprising that I'm having such problems right out of the gate.

Screenshots:

Black screen on boot up:





HD activity lights flashing:

Comments

Post a Comment

Popular posts from this blog

Team Conflict Avoidance System Erratic Behavior

The below video shows me discovering conflicting games that were auto-scheduled by the StackSports scheduling system despite a team conflict rule having been previously created to prevent this. Usually the system adheres to the team conflict rules I create. Sometimes (inexplicably) it doesn't. This video is an excellent example showing this is actually a bug, and not intended behavior. After making a subtle adjustment to a game time, it's like the existing team conflict rule "woke up", and triggered the red alerts it should have done all along. In fact it should have scheduled these games in a way to avoid the conflicts in the first place, and therefore the alerts wouldn't have ever had need to exist. This video shows problems such as: Conflicting games that should never have been scheduled at the same time (based on the team conflict rules that were already defined). Missing alert/warnings for "error" games that have problems (i.e. the gam...
Read more

No grace in verbatim communion prayers

Image
I do not appreciate the intolerance for minor grammatical errors in the spoken (typically ‘read’) communion sacrament prayers when recited by the LDS young men during congregation worship services. I understand the covenant nature of the prayer, and thus the importance of the meaning of the covenant being accurately conveyed in the prayer. But the intolerance for minor nuanced accidental variations that do not change the details of the covenant or prayer seems nonsensical. It is disruptive to the service to require the young man to repeat the entire prayer for such an error. And it potentially embarrasses the young man unnecessarily. Considering the degree of grace and leniency church members seem to grant Joseph Smith for much more significant variations in details he communicated (such as stating he was 14 years old, and also that he was 16 years old, when he had his first vision), I don’t understand why we can’t grant our young men a bit of grace for a much-less significant nuanced...
Read more