Hey everyone,

 

Running through my syslogs, I came across this line multiple times (example pulled from syslog.2.txt):

 

Apr  8 23:53:22 Smalls kernel: ata1.00: Enabling discard_zeroes_data
Apr  8 23:53:22 Smalls kernel: sd 2:0:0:0: [sdj] Attached SCSI disk
Apr  8 23:53:22 Smalls kernel: ata14: SATA link up 1.5 Gbps (SStatus 113 SControl 300)
Apr  8 23:53:22 Smalls kernel: ata14.00: ATAPI: MARVELL VIRTUALL, , 1.09, max UDMA/66

 

So my concern is 2 parts:

 

1) Does that mean sdj is the device that is only pulling the 1.5Gbps link? If so, that is my cache drive

2) It mentions MARVELL below it (hisssssss), does that mean it is on the MARVELL sata port? I double checked it compared to the board and the cache drive is plugged in to SATAIII_0 which should be the Intel SATAIII port.

 

 

 

Potential Improvements:

Currently I'm running LSI 9210-8i x3 cards (IT Mode) in PCIE5/PCIE4/PCIE3. I was looking in to getting a RAID expander card and saw an ebay posting for: (E91267 E91267-202 E91267-203 INTEL RAID EXPANDER STORAGE CONTROLLER). So, a couple questions came from that:

1) Does the version of the card matter for unraid? E91267 vs -202 vs -203

2) Does anything extra need to be done to the card to get it to work with unraid, like something similar to IT Mode or a specific firmware?

 

^ - Unfortunately I have never used a RAID EXPANDER card before, so that drives additional questions:

1) Does a card like this require an additional power connector?

2) Does the card work like this: I plug in my 5 bays in to 5 of the connectors on the card, then the 6th connector goes to the LSI 9210-8i via mini-sas to mini-sas cable? If that is the case, does the cable location matter on the expander card?

 

Thanks in advance for your time! Diagnostics attached

 

 

smalls-diagnostics-20190604-1904.zip

Edited June 5, 20197 yr by GroxyPod
Corrected wrong LSI Card #

1 hour ago, GroxyPod said:

1) Does that mean sdj is the device that is only pulling the 1.5Gbps link? If so, that is my cache drive

Yes, 860 QVO 2TB

 

1 hour ago, GroxyPod said:

1) Does a card like this require an additional power connector?

It have 2 power source, PCIe or Molex, either one would be fine.

 

1 hour ago, GroxyPod said:

If that is the case, does the cable location matter on the expander card?

Not sure.

 

If connect 20 disks, only 1 port remain for HBA, so not enough bandwidth.  But if dual link to HBA then only could connect 16 disks.

You may need 2nd HBA or connect remain disks to onboard SATA or you need found 36 port Expander.

 

BTW, does any bottleneck in current setup. i.e. when parity start, max speed got ?

Edited June 5, 20197 yr by Benson

Thanks for looking into this Benson, I appreciate it!

 

25 minutes ago, Benson said:

Yes, 860 QVO 2TB

 

I figured that was the culprit. I went ahead and swapped the SATAIII port from SATAIII_0 to SATAIII_1 to see if that improves anything. Keeping an eye on it for possible bad port or cable. Booting up UnRAID did not produce the issue that was occurring above so we shall see.

 

22 minutes ago, Benson said:

It have 2 power source, PCIe or Molex, either one would be fine.

 

Fantastic. I'd prefer to just use PCIe power if that's the case anyway. Less cables to run across the board.

 

22 minutes ago, Benson said:

Not sure, if connect 20 disks, only 1 link remain for HBA and not enough bandwidth, so best have 2 link to HBA and connect 16 disks.

Then you also need 2nd HBA or connect remain disks to onboard SATA.

 

Gotcha. So it sounds like I'd be stuck with a 3 card setup regardless.

 

28 minutes ago, Benson said:

BTW, overall seems haven't benefit, extra cable need. I think you shoud check does any bottleneck in current setup first. i.e. when parity start, max speed was ?

 

My last parity run (06/01/2019) was: Duration: 23 hours, 59 minutes, 35 seconds. Average speed: 115.8 MB/sec. When Parity starts, its usually in the 160 MB/sec range until it reaches the edge of one of the disks. I've got a mixture of 4TB/6TB/8TB disks in the array at the moment.

 

As far as bottlenecks go, I'm not necessarily experiencing any other than an issue with slow DB messages on the Plex docker. Given that the cache drive was the subject to that error, it could of been related. I will continue to monitor and work on updating my disk tunables if that turns out not to resolve the issue.

 

 

 

2 hours ago, GroxyPod said:

Does that mean sdj is the device that is only pulling the 1.5Gbps link?

No, ATA14 is the Marvell Virtual adapter, not a drive, and that link speed is normal, just make sure in the future you don't use none of the Marvell ports on that board, the first 4 white ports, as they are a known problem and tend to drop disks, you're not using them currently.

2 minutes ago, johnnie.black said:

No, ATA14 is the Marvell Virtual adapter, not a drive, and that link speed is normal, just make sure in the future you don't use none of the Marvell ports on that board, the first 4 white ports, as they are a known problem and tend to drop disks, you're not using them currently.

 

Thanks johnnie.black!

3 hours ago, GroxyPod said:

My last parity run (06/01/2019) was: Duration: 23 hours, 59 minutes, 35 seconds. Average speed: 115.8 MB/sec. When Parity starts, its usually in the 160 MB/sec range until it reaches the edge of one of the disks. I've got a mixture of 4TB/6TB/8TB disks in the array at the moment.

WD RED 4TB 5400rpm disk mark, so speed normal.

 

 

Thanks @johnnie.black point out the mistake.

9 hours ago, Benson said:

WD RED 4TB 5400rpm disk mark, so speed normal.

 

Awesome, thank you!