LSI M1115 in DS380 (or which Marvell controllers don't puke in 6.x)

[tucansam]

I am replacing a four-port Marvell controller in my Silverstone DS380-based unraid 6.x server because it is having major issues with 6.x

 

I ordered an M1115, and it appears that it will not fit.  I can try a PCIe ribbon extender but that will require major case mods.

 

So two main questions: is there any Marvell chipset that doesn't completely fail under unraid 6.x, and, failing that, are there any worthwhile 4-port (or 8-port if they are SHORT) controller cards that aren't Marvell based and still work with 6.x?

 

Thanks.

 

EDIT

 

Or, failing all of the above, any idea as to when the Marvell chipsets will become reliable under unraid again?

[Fireball3]

The only short 8-port card I have seen so far is the Highpoint RocketRAID 2720SGL.

Unfortunately it's a Marvell also.

 

You might be able to find some short generic 4-port cards with JMB or ASMedia chipsets!?

 

[ashman70]

While its only four ports, I have one of these cards in my DS380 so I know it fits, and I am running 6.3.1 no issues.

 

https://www.newegg.com/Product/Product.aspx?Item=N82E16816124064&cm_re=four_port_sata_controller-_-16-124-064-_-Product

[tucansam]

On 2/27/2017 at 7:39 AM, ashman70 said:

While its only four ports, I have one of these cards in my DS380 so I know it fits, and I am running 6.3.1 no issues.

 

https://www.newegg.com/Product/Product.aspx?Item=N82E16816124064&cm_re=four_port_sata_controller-_-16-124-064-_-Product

 

Thanks, but that's a Marvell chipset from what I can tell.

 

The M1115 will not fit my DS380.

 

Are *all* Marvell chipsets broken under 6.x?  I need a short 4-port (minimum) card that will work under 6.x.  I have a Supermicro in one server (it also will not fit the DS380) that, so far, has not had issues.  The Marvell I have in my now broken server was working fine until a few days ago (literally nothing on the server has changed except disks getting more full) and now I am being told Marvell has issues with 6.x..... So what's the alternative for those of us with small server cases?

 

Thanks.

[Fireball3]

4 hours ago, tucansam said:

The Marvell I have in my now broken server was working fine until a few days ago (literally nothing on the server has changed except disks getting more full) and now I am being told Marvell has issues with 6.x.....

I'm not sure how reliable this statement can be!

Have you done some testing to prove it's

a. not a defective controller ?

b. an unRAID 6.xxx issue ?

 

There are many Marvell-based controllers out there and if it really was an unRAID 6.xx issue, the forum would probably suffer a DDOS.

[tucansam]

I agree.  I figured there would be 1,000 threads about Marvell if 6.x broke them 100%.  From the reading I've done it seems sporadic, and only affecting some users.  Also very little testing and troubleshooting seems to have been done, maybe I can help fix that down the road.

 

This system has been balls-dead reliable from day one, whereas my other unraid server has had issues since the day it was built.  I was hoping for the fastest fix possible to get it back to "five nines" uptime, and the LSI seems like it will fit the bill.  The last thing I want to deal with is driver or chipset issues, and the research and troubleshooting that would come with them, with a server I use every day. 

 

I have a Supermicro (Marvell) in another 6.x system that (as yet) hasn't failed.  And I think I figured out how to get the LSI I just got to fit my DS380, so I am hoping to get that in this weekend.  Being without the server for a week as been a major PITA.  

 

Once I get the system back up, I plan on putting the Marvell controller in a test system and doing some more troubleshooting.  If it turns out to simply be a failing card, at least I'll have the speed upgrade and double port capacity of the LSI.  I had planned on moving some dockers to a disk outside the array anyway, so now I'll have that option.

[tucansam]

Well I shoe horned the controller in, and the array came up.  Had to do a new config, and am now syncing parity.

 

Speeds are 8.9MB/sec.   <-- not a typo

 

What could be causing these horrible speeds?

 

The cable I ordered is rated for SATA 3, and the controller is set to defaults.  I entered the config utility and there are no options that looks like they would affect speeds.

 

ETA now @ 6.1MB/s

[Fireball3]

Now this is the Marvell controller from your other, non failed, system?

 

Can you give some details on your hardware?

CPU, Board, drivecount

Maybe a diagnostics file?

 

Btw, what are the major issues you mentioned in your first post?

[tucansam]

This is the LSI M1115 that replaced the Marvell controller I was having issues with (Disk 1 was dropping all the time past week or so)

 

Asrock 1155 ITX board, 8GB RAM, i5 2500, 8 drives, 8TB WD Red Parity and the rest a mix of 2, 4, and 4TB WD Reds and Greens.  

[Fireball3]

What is the health status of this disk 1?

Have you considered rebuilding the array without it?

At that rate I wouldn't go on with the sync!

[tucansam]

Disk 1 passes all SMART tests, so far it looks like it was a controller problem.

 

 

[Fireball3]

I have serious doubt's that the M1115 has the same problems like the Marvell!

Something is slowing the whole thing down!

Is anything working in parallel on the drives? Another task/process?

Have you double checked the cables?

 

 

 

[Fireball3]

17 minutes ago, tucansam said:

Disk 1 passes all SMART tests, so far it looks like it was a controller problem.

 

 

I would try a new config without this disk 1.

There is nothing to loose.