All,

I've noticed that over the last couple months my UnRaid writes have slowed way down.  I have it configure with a cache drive.  I used to get ~20MB/s kind of write, but now I"m at 2MB/s.

 

I have not changed anything in over a year on the server, just let it run along happily.  I'm running version 4.4.2.

 

The UnRaid is connected to the PC via a GB switch so I don't think it is that.  It has never been the bottle neck before.  I've tried writing to UnRaid from 4 different computers and they all experience the same issue.

 

I don't see any errors or warnings in syslog.  Though I'm not positive I know what to look for.

 

Hoping someone can help me figure out why performance went wrong.  It takes longer to copy the DVD than to watch it

 

thanks,

dave

Ok of course I see something after I post

 

I'm actually running BubbaRaid and was looking at the System information page.

 

It showed the memory at 99% utilized so I logged in and ran top.

 

I'm seeing 34% wait state.  This is why I"m slow.

 

I then sorted procs by mem used and the largest one is smbd at 2% of the memory.  There are no processes using any memory.  My uptime is 150 days, but it seems that there is a memory leak somewhere.

 

Guess a reboot is in order.

 

 

Ok me again after the reboot.

 

After reboot memory was at 53% so I started the file copy again.

 

Speed is 1.65 MB/s memory is at 97% used and I"m hitting hith wait states again 8-34%.

 

I"m out of ideas.

 

 

logfile attached

syslog.txt

Your syslog looks very good.  I found your syslog from a year and a half ago and compared the 2, and found it essentially identical, apart from the addition of 3 drives.  All drives are setup correctly, without any issues, and at the best speeds your system is capable of.  The network connects at gigabit speeds, without any apparent issues.  Assuming that the period covered by this syslog included some of those slow transfers you mentioned, then there is no explanation available yet.  You only have the minimum amount of memory (512MB), so that is why you were using all available memory.  Most have 1GB or more, and unRAID tends to allocate the first 900MB for LowMem and any additional for HighMem, using the extra low and high memory for caching and other app usage.  Since you are starting out with much less, all of it becomes LowMem, and any unused from that is used for caching (I believe, I'm not a Linux expert).

 

I can suggest a couple of things you can try, to narrow the problem down:  first get SMART reports for all drives.  Perhaps one of the drives is starting to fail, and media issues are causing delays.  Then you can try some read and write tests to isolate which drive or drives are performing abnormally.  Copy a gigabyte file to each of the data drives and the Cache drive (but not the User Share system), and time how long it takes.  Then copy that file back, and time how long that takes.  That will give you some rough read and write measurements for each drive, and may help you discover 1 drive with much worse numbers.  To test without the router or switch, Telnet/PuTTY into your server and use MC to do the same read and write tests between the data and Cache drives.  These tests do not directly test the parity drive, but if write performance is drastically less than read performance, you will have a strong indicator of a bad parity drive.

Rob,

 

Thanks so much for looking into this.

 

I copied a 1GB file to each of my drives and all copied at around 10-12MB/s.  This is much better than I was seeing before so I looked into it a little more and found that the issue was user error.  When I measured before I was not using copy, but had a program copying for me.  It turns out that it is the program limiting the copy speed not UnRaid.

 

Marking as solved.

 

thanks again,

dave