The problem.  If I try to do normal read/write operations that I have been doing for years, my UnRAID cannot keep up with the transfers.  So video freezes or slows down that I am watching. 

 

Okay, for a brief history.

 

I first went unRaid 5 years okay.  Ever since, there has been a basic daily ritual.  My peak use period is in the evening.  During  this time I have:

 

• 5 Security cameras streaming video and recorded on unRAID (writes 6GB per hour, or 1.7 MB/s) directly to disk1, bypassing cache drive)
  
• One Blu-ray rip streaming (average 25 Mbps transfer; peak 40 Mbps)
  
• One DVD rip streaming (average 6 Mbps transfer; peak 10 Mbps)
  
• Batch of files copying to UnRAID (65-70 MB/s write to cache, ~50 MB/s read when CRC checking file transfers)
  

 

This was literally a daily routine, writing 70 MB/s while reading 56 MB/s (peak).  This is all on a Gigabit network, with good equipment that all tests/functions near max levels (over 100 MB/s reads/writes achieved when system was deployed). 

 

I have had this basic routine for 5 years, with no problems until recently.  In September 2015, I migrated to UnRAID 6.1.2; installed a 4 TB parity drive and my first 4TB data drive.  After the migration, no problems to speak of.  In November, I went to 6.1.3, and again no problems that really stood out that I recall.  By the end of December, I had 4, 4TB drives in total, 15 2TB drives, and a small 512MB cache drive.

 

So in December, I don't recall having any problems with doing the above, simultaneously.

 

At the end of December, I built a 2nd unRAID and offloaded all the security cameras to the 2nd unRAID so that constant 24/7 write goes away on my data unRAID.  So you would think that would improve conditions, not worsen them.  In January, I started migrating to XFS from ReiserFS.  Today, all but 4 disks are converted.  I also replaced one more 2TB disk with another 4TB disk, to give me more space for migration.  Last month I upgraded to 6.1.9.

 

It is around that time I started noticing some problems.  My NIC started reporting overrun errors that I hadn't seen before.  I started getting some audio dropouts on video playback.  And if I try to play the two videos and copy files at the same time, the video will freeze (unRAID can't keep up with  the data transfer requirements).  Pause  the data transfer, and playback instantly goes back to normal.

 

I've even tried having my PC connect with a 100 MB connection instead of gigabit, with  transfer speeds limited to 15 MB/s read or write, and it still causes video to freeze throttling it back that much (that should just be a read of 21 MB/s out of unRAID).  And when I check my unRAID CPU utilization, it is never more than 50%.

 

So I am at a loss for the cause.  At first, I thought it was the Realtek NIC.  No idea when it started getting all the Rx overrruns or if that was caused by some driver change from 5.X to 6.X, or something from 6.1.3 to 6.1.9.  It used to not always be that way.  So I disabled the onboard NIC and installed an Intel Pro 1000/GT.  No more overruns, and audio dropouts seem to have gone away.  But video still freezes when trying to do the above.

 

I will say that it is a PCI card I had lying around, and I have an Intel PCI-E card on the way.  PCI bus is 133 MB/s, but it is shared.  My MB only has one PCI slot, so this 133 MB/s isn't being shared by anything external, but maybe onboard audio/video is consuming some of that 133 MB/s bus.  I don't know.  It seems like PCI bus should have plenty of bandwidth to support this.  The PCI-E adapter will have a dedicated 250 MB/s available to it, so if the bottleneck is there, the card will take care of it.

 

Could it be something to do with XFS?  Could XFS be causing some type of bottleneck on the data transfer out of  the HDD's.  My file transfers are to cache drive, outside array.  My video playback is usually from two different drives in the array.  So no one drive is being hit will all the requirements, so I wouldn't think it would be anything at drive level.  But could it be file system operations and the unRAID level?

 

Could there have been changes to unRAID between 6.1.3 and 6.1.9 that made it less efficient at data transfer?

 

I'm really not sure if I should be looking at rolling back to 6.1.3, or migrating back to reiserfs, or what to do next?  I would like to get back into my old routine though.

 

 

 

 

Rolled back to 6.1.6 and nearly all my problems have gone away.  Still having the issue of RX Drops on my NIC's, both onboard Realtek, and Intel Pro 1000 GT.

 

Oddly, the 1000 GT had zero drops over three days on 6.1.9.  It had 720 drops after about 4 hours last night.  I also put the NIC in promiscuous mode, since some drops may be normal.  They were still there, so it seems to indicate they are actual drops.

 

My only other concern with this NIC is that my throughput looks like a since wave now from 40 MB/s to 100 MB/s, averageing about 65-70 MB/s.  Back on 5.0.6 (last time I liked), it was just continuous throughput with very little fluctuation.  That was on the Realtek NIC, so don't know if t his is some function of the Intel NIC, something weird in the unRAID version, or what.

 

I also tried an Intel Gigabit CT and it was getting thousands of drops per second, just like the Realtek NIC.

 

I don't know what it will take to sort the network out after going 6.1.X...