Writing to user share fails, wrongly reports lost connection

[gordongraham]

Hi

 

I have recently put together a fileserver using Unraid 4.7 Plus as the OS. It is an HP Microserver N40L with 2GB RAM, populated with 5 drives (3 x 2TB, 1 x 1.5TB, 1 x 1TB Samsungs, one of the 2TBs is parity). Everything works great, all my data copied over fine initially writing large chunks manually to each of the 4 disk shares.

 

I set up 2 user shares, each exported to read/write (one Movies, one TV). Sometimes I can copy files directly into the user shares without any problem, and sometimes it just hangs at "calculating" and times out after about 60 secs with something like "//myserver cannot be found, please check it is connected to the network" (I am at work just now, so cant remember this message verbatim). It is random when it happens too, ie not just with large transfers. It happens to both the tv and movies user shares (split level 2 for each, dont know if thats relevant). It also seems to be irrelevant whether or not the disks are spun up at the time transfer initiated.

 

After the transfer fails, the web ui is still visible, shares still accessible, all seems perfectly connected and working as normal.

 

Each disk has at least 400GB free space. I have tried rebooting both ends, same result.

 

I googled for a solution, but couldnt find much. One forum post from 2007 suggested it could be because writing to a user share essentially copies to RAM first and then to the array, but I couldnt see how this would still be the case, if it ever was.

 

Any thoughts would be much appreciated. If you need any further info let me know. I will grab a copy of the syslog and include it here this evening.

 

Cheers

Gordon

syslog.txt

[Johnm]

A syslog would be helpful.

 

Otherwise we are just guessing without much to go on.

 

Few guesses.. although probably not the cause.

your drives were asleep and windows didn't feel like waiting for them to spin up..

You have DHCP timeout to short and the IP changed. windows was trying to find the box? It was not fast enough..

also, if you are on vista, this is a known bug in vista that had 2 hotfixes that didn't solve the issue.

 

if you retried the copy right away did it work?

[gordongraham]

Hi John, thanks for the reply. I have now attached the syslog, though it seems to me that it only includes very few entries - the 2 timestamps dont seem to cover much ground. Have I done something wrong? A copy directly to a disk share failed this morning at around 8:20 (roughly), right before I copied the text from the syslog, so i assumed that would be included but I cant see it.

 

The drives are usually asleep when copy/move initiates but seem to wake fine. The failed "calculating" dialog appears sometimes even if all drives are manually spun up directly beforehand.

 

The box has a static ip reserved in the router so shouldn't change - I have checked this from time to time and it does seem to stick so I doubt DHCP timeout is the problem.

 

I am on Windows 7, not Vista.

 

If i retry the copy/move straight away after the failure, it does work.

[gordongraham]

Can anyone have a quick look at the syslog and tell me if its complete? I have looked at a recent one also and cant see any entries for copy/move operations. I suspect im getting the wrong one from //myserver/log/syslog

 

Thanks

[gordongraham]

Second syslog attached.

syslog2.txt

[RobJ]

Both syslogs look fine, no errors at all.  The second log is after the system had crashed in its previous run, so transactions are replayed and a parity check is performed.

 

You have a gigabit NIC in your server, but are only connecting at 100mps.  If that is what you are expecting, then no problem, but if otherwise, then your transfers are much slower than necessary.  DHCP looks fine.

 

I have heard of others with occasional Windows transfer timeouts, but can't help you myself.  Hopefully others with more experience in that will pop up here.

 

I've never heard of the "calculating" dialog, sounds like a Windows issue, is not an UnRAID thing.

[gordongraham]

Hi Rob, thanks for your reply. I am running through an older router so although both ends have a gigabit NIC, the integrated 10/100 switch in the router is a bottleneck. New router on its way.

 

I have had little joy solving the "calculating" problem. It still happens with almost every transfer. I suspect it may be a driver issue with my PC's D-Link Gbit NIC. Ill try a different driver and see if that fixes things.