Losing network whilst in mid-transfer

[pras1011]

Hi Tom.

 

I think the problems are due to a lack of space on the server.

 

I have 12.7TB on my server with only 2.1TB free (according to Windows 7). I set the allocation to High Water. The first attempt at a transfer (writing to the unraid server) fails but I noticed that one the hdd's activity light on the server is on now continuously on for a very long time. Checking the main unraid page shows that the hdd is continuously reading. Maybe its trying to calculate the space required for the file?? When it finally stops I then try the transfer again and it works. The Windows "discovering items" is a bit slow but it works. The same thing however happens in Most Free.

 

I then deleted some files until there was about 3.1TB of free space on the server. I then did a transfer and it works perfectly first time!!!! The file was "discovered" by Windows instantaneously. When the free space reached 2.54TB it went back to this annoying problem.

 

One way around this was to set the allocation method to "fill up". There is still a long hdd pre-read though. The "discovering items" is still a bit slow but it works.

 

I put then put 2 x 2TB drives in the server which now totals 18TB with 5.6TB of free space. The server has now started responding normally again. Transfers were instantaneously "discovered" without the long pre-read of the disc. There are no syslogs regarding this. And it doesn't seem to matter if the file is 444kb or 40GB.

 

Any ideas Tom?

[ohlwiler]

This is a long time and known behavior of unRAID. With slow, large hard disks that are nearly full, the initial transfer will fail, while the second transfer will succeed. See here:

 

http://lime-technology.com/forum/index.php?topic=6098.msg58633#msg58633

[pras1011]

Thanks for that!

 

The drives that are nearly full on my server are 7 x 2TB Samsung F4 drives. These are 5400 rpm. I have 7 x 2TB Hitachi 7200rpm drives to fill now. Maybe I could replace the F4 with the Hitachi?

 

I have asked aobrien to send me the fix he applied to sort out his problem.

 

But surely there could be a fix that Tom could introduce to solve this annoying problem?

[pras1011]

I have found something odd. I have set the allocation to Fill Up.

 

When I try a transfer the problem occurs and the transfer fails. I wait for the target disk to stop reading.

 

I then try the transfer again and it works without delay. I then try again and works perfectly. I changed to different allocation methods and it works perfectly!

 

I then then reboot the server and the problem occurs on the first transfer but works again there after.

 

Confused!!!

[BRiT]

My completely wild ass guess is this has to do with how ReiserFS handles file table expansion when the disk is nearly full and shfs or samba timing out. The first attempt takes quite a bit of time for ReiserFS to do it's internal house-keeping while the second or more attempts don't need the housekeeping performed.

[pras1011]

Maybe Tom could step in a provide a response?

[pras1011]

I made a mistake. The reason why its working is because the free space is currently over 2.54TB.

 

The problem exists when the free space drops below this so therefore I will have to wait till then. But I will fill the slower Samsung F4 drives and then I will move to the faster Hitahci 7K2000 drives to see if that makes a difference.