I have been running a gen1 unraid box with the original Intel board and 12 300gb IDE drives for several years now.... I recently put together a new box to replace it based on Abit AB9pro, Q6600, 8gb DDR2-800, 2 SIL Image PCIe cards and an Adaptec 4 port card, the intent being I have room for (eventually) 12 1tb drives for the array, a 500gb Cache drive and a Linux boot drive currently 80gb, AND each drive has it's own PCIe channel.

 

I put the box together a few month back with 6 1tb drives, the 500gb cache and 80gb boot drive.  Parity builds were VERY fast, and drive to drive transfers over ethernet from a windows vista sp1 box (parity disabled) were close to 30 MBytes/sec and 85% Network utilization.  Then I got into other things and it languished.

 

Sunday I dismantled the old server and prepared to transfer the data from the old drives to the new box.  I hooked up one of the data drives with an IDE/SATA converter board and prepared to plug it into the ESATA port.  First, I booted up a live distro disk of Slax, a slackware derivative.  I opened a window showing the online storage, five 1tb drives (I removed the parity drive), one 500gb drive, and the 4 partitions of the 80gb drive.  I plugged in the ESATA cable and the 300gb drive showed up and I could browse it.... so far so good.  I rebooted into unraid and stopped the array, added the 300 gb drive as disk 6 and re-started the array.  From my vista box I could browse the drive and see the files... good again!  I opened a window on disk 6 (the 300gb) and another on disk 1 (a 1tb).  I started copying a folder of files and noticed it was going quite slow... about 2.5 MBytes/sec and 7.5% Network utilization !!  Yuck....

 

At this point I had to go to work, so later in the morning I remoted into the vista box and tried again... same thing, 2.5 MBytes/sec.  I stopped the array and rebooted the unraid box.  Tried again and Whoa!  25-30 MBytes/sec and 85% Network Utilization.  Cool !!!  I chalked it up to a typical "needed a reboot" and went on copying files (notice the big ommision here?  I should have done a syslog!!!).  Later, I logged in and noticed that the connection to the unraid box was lost.... hmmmmmm  ?

 

When I got home, I saw the gigabit switch which both machines are plugged into had a yellow light meaning 100mhz connection to unraid.  I shut down and re-booted the unraid box from the console and replaced the cable to the switch since it was a new one I had never used before.  Got everything back up and boom.... back to 2.5 MBytes/sec... Damned!

 

I tried everything I could think of, including starting from scratch and unplugging the drive and booting slax and going through the whole procedure... no dice, same poor performance.  I'm stumped.  Syslog attached.

 

 

I had the same issue as you (except I had bought a new Gigabit switch).. It turned out to be a problem with Vista in the end and it was resolved by rebooting the Vista machine..

 

The other thing I did was to follow this guide: http://lime-technology.com/forum/index.php?topic=1247.msg8332#msg8332

 

Make sure you have SP1 installed as well.

How odd.... one time I have to reboot unraid, another time the windows box....

 

I finished copying the contents of disk 6 and shut down unraid, replaced the 300gb disk 6 with another drive.... booted unraid, did a restore (  ) to set up the new disk in the array and went to windows to copy files... ugh... 2.4 MBytes/sec....   rebooted windows and voila... 27 MBytes/Sec.... weird 

 

 

Reason #99 why I won't downgrade to Vista.

 

 

Bill

Has to be more to it than Vista... I say that because the first time it was unraid that had to be rebooted...

 

 

Could it be some type of TCP window definition?

I remember there used to be a tuning program for XP that allowed adjustment of segment and window size.

 

 

Perhaps try it with a cross over cable and take the switch out of the equation.

 

 

Also keep in mind, when the buffer cache gets full, network traffic slows down while the cache is flushed.

 

 

Any TCP Errors show up in ifconfig ?

 

 

Could it be some type of TCP window definition?

I remember there used to be a tuning program for XP that allowed adjustment of segment and window size.

 

 

Perhaps try it with a cross over cable and take the switch out of the equation.

there's an idea.... my original tests were with a direct cable connection... good call 

 

 

Also keep in mind, when the buffer cache gets full, network traffic slows down while the cache is flushed.

 

 

Any TCP Errors show up in ifconfig ?

None... everything is clean...

 

 

I've read through several times, and it still sounds like you are adding these data drives directly to your unRAID server.  If not, I apologize, and you can ignore this post.  If so, why not take the Windows boxes and the network out of the loop and just do disk to disk copying?  You have already taken the parity drive out of the loop.  Use Telnet/PuTTY to control Midnight Commander, and get the fastest copying of all.  mc is now included in unRAID.  See http://lime-technology.com/wiki/index.php?title=Transferring_Files_Within_the_unRAID_Server.

 

Edit:  I could not read your attached syslog.rar.txt, possibly corrupt?

 

Edit again:  I know you are quite knowledgeable, too knowledgeable not to copy directly, so I probably missed something simple...

in a nutshell, I tried it ... but I have many terrabyters of music files and many have been tagged with filenames with odd chacters, such as "Blue Oyster Cult" where the y has two dots over it.... these seem to really foul up in raw linux... using MC they often end up not being copied, or worse, being coppied over each other where I end up with one file of garbage...