The specs of my Unraid box are as follows:

 

ASUS P5B M/B

8GB RAM (2x 4GB DDR2 800)

Intel Core 2 Quad CPU

1x IBM M1015 Card

6 WD 2TB Green Drives

1 Samsung 2TB drive

1 WD 1TB Green Drive (Cache)

1 WD 3 TB Red Drive (Parity)

Unraid 5-RC12a

 

Parity and Cache drives are connected to onboard SATA ports, the other drives are connected to the M1015 card.

 

I find that when Unraid is doing a parity check it averages around 28MB/s, takes about 28 hours to complete with an average of 35MB/s. As other people on these forums appear to be getting speeds significantly faster than I am with similar hardware or not much better I have been trying to diagnose why mine are so slow.

 

I initially thought I had it cracked when I discovered the PCIE slot I had the M1015 card plugged into was x4 and not x8 so I moved it to the PCIE x16 slot. I expected big numbers when I booted it up again and kicked off another parity check only to find it was running at the same speed, zero performance improvement. Now that I think about it, I didnt really see any improvement moving my drives from 2 generic PCI SATA cards to the single M1015 card.

 

I checked the syslog for errors and nothing is appearing to indicate a fault with a particular drive that may slow things down, is there another log file I should be checking?

 

Is my hardware bottlenecked somewhere or are these speeds roughly what you would expect for the above hardware configuration?

 

Any advice appreciated.

 

Thanks! 

That's a fairly dated motherboard with PCIe v1, but that still provides 250MB/lane, so an 8 lane card should provide 2GB of bandwidth -- more than enough for the 7 drives you have connected to it.

 

What are your tunable disk parameters set to?  [settings - Disk]

The md_num_stripes should be >= the sum of the other two.    I'd try 2048/768/1024 and see if that improves your parity check times.

 

If that doesn't help, you'll want to get SMART reports on each of your disks, and run a performance test on them as well.  You may have one disk that's causing problems ... and a parity check is limited by the performance of your slowest drive.

 

I have not touched the disk parameters they would be set to whatever the default is. I'm at work at the moment, will check out the values tonight when I'm home again and try the ones you suggested.

 

If I have 1 disk that has an issue would that throw out the whole parity check process or only affect it during the check of that one drive?

 

Thanks for your reply!

If I have 1 disk that has an issue would that throw out the whole parity check process or only affect it during the check of that one drive?

 

One disk will slow down the entire parity check.  The disk are all read at once, so the longitudinal parity bit can be computed and checked.  This cannot happen any faster than the speed of the SLOWEST disk.

 

Since you have 7 data disks plus parity, but the data disks are all 2TB, then after the check passes the 2TB point, the only disk involved is the 3TB parity disk.  Pay attention to the speed it's showing just after it passes the 2TB point.

 

OK, will get a SMART report for all of my disks, what variables should I be looking for to establish if there is an issue?

Changing the disk settings to ones specified by Garycase didn't make any difference to the parity check speeds, still hovering around 30MB/s. Do I need to stop the array or reboot the server for these to take effect?

 

According to the smartctl command none of my disks have errors. One of them had a single reallocated sector, the rest were at zero. One of them had a high UDMA CRC Error Count which red balled the drive a few months ago but I resolved the issue by replacing the SATA cable.

 

I am due a hardware upgrade for this box, may just go ahead with that, add each drive to the new box and monitor the parity check speeds to see if they drop off.

 

Thanks.

Changing the disk settings to ones specified by Garycase didn't make any difference to the parity check speeds, still hovering around 30MB/s. Do I need to stop the array or reboot the server for these to take effect?

 

According to the smartctl command none of my disks have errors. One of them had a single reallocated sector, the rest were at zero. One of them had a high UDMA CRC Error Count which red balled the drive a few months ago but I resolved the issue by replacing the SATA cable.

 

I am due a hardware upgrade for this box, may just go ahead with that, add each drive to the new box and monitor the parity check speeds to see if they drop off.

 

Thanks.

 

That's actually a good idea ... EXCEPT if you need to keep the data on all the drives, you can't add them once you've assigned a parity drive on the new array (it will clear the drive).    What you CAN do is add a drive;  assign parity and note the time;  then do a New Config WITHOUT a parity drive;  add another of the old data drives; then assign parity again;  note the time;  do another New Config without parity; etc. ...

 

Thanks for the advice, i'll look at doing that.