2

In one of our offices we have a Dell PowerVault 745N NAS device which acts as the main file server. Its running 32bit Windows 2003 Storage Server SP2 with 3GB RAM. The server holds around 60 users HOME folders, which are mapped via AD.

The office clients are a mix of XP SP3, Vista and Windows 7. Occasionally the server will completely hang when transferring large files. When the hang happens the console becomes unresponsive with only the mouse active and blank wallpaper. Sometimes stopping the copy frees the server, sometimes not.

The hanging can last around 20 minutes. During this time other servers also become unresponsive with blank wallpaper at the console. If you do manage to get onto another server the taskbar and run commands are unresponsive.

This also transcends to the client computers sometimes with explorer crashing. I'm guessing this is due to the HOME folder mapping.

Eventually the NAS server with free up and everything will be back to normal.

The server is configured as follows:

PERC 4/DC DATA 2 - 12 SCSI HDD - RAID5

SHADOWCOPY 2 SCSI HDD - RAID1

CERC SATA DATA 11 4 SATA HDD - RAID5

OS 4 SATA HDD - RAID5

All the drivers and firmware is up to date. I've been through all the diagnostics with Dell and the hardware has come up clean including full HDD tests on the arrays. The server has NOD32 installed as the AV, but the hanging happens when it is uninstalled.

There are no errors in the event log when this happens and we don't have any errors logged on any of our ProCurve switches. DNS is fine on the domain and AD from what I can tell is running happily.

There are no DFS or NFS shares setup either. All the shares are standard Windows.

I've unchecked the allow the computer to turn off this device to save power box under Power Management on the NIC.

"Set Link Speed and Duplex to Auto-negotiate 1000 " Increased Receive Descriptors buffer from 256 to 352 (reserves more CPU resource for handling data)

I've run network traces using network monitor and have found the following: 417 8.078125 {SMB:192, NbtSS:25, TCP:24, IPv4:23} 192.168.2.244 192.168.5.35 SMB SMB:R; Nt Create Andx - NT Status: System - Error, Code = (52) STATUS_OBJECT_NAME_NOT_FOUND

I've tried different cabling; NICs and switch ports all with the same result. Transferring files from other servers on the domain is fine. All I haven't done is run CHKDSK on the drives to look for any file system errors.

On the Vista clients I have also run netsh interface tcp set global autotuning=disabled with no result.

Could it be that the server has a faulty drive or that the I/O is too much for it to handle?

Any ideas why would the hang cause issues with the other servers on the LAN?

Many Thanks.

1
  • How big are the files?
    – jftuga
    Jan 3, 2011 at 0:51

4 Answers 4

1

Have you watched the server's memory usage while copying those large files? Personally, I've found that moving a 10GB file is a great way to crater my windows 2003 server... Have a look at this topic: Windows Server 2008 x64, Large File Transfers, and Memory Usage

Some alternative file copying tools are discussed for handling large files. I've been having good luck with RichCopy.

0

This sounds like copies to the server are outpacing the disk subsystem's write capabilities (or part of it). The 12xSCSI Raid 5 group on the PERC controller should be able to support large sequential writes at >200Megabytes/sec (hopefully a lot more) but the SATA group on the CERC probably struggles to get past 70Meg/sec and might even be quite a bit slower than that. A large copy targeted at the SATA Data Raid pack could easily come in over a GigE connection faster than that and if that does happen then Windows 2003 will consume local as much local memory on the server as it needs to buffer the copy. The buffer will grow at the expense of everything else on the server, it will even cause core OS services to page out for example which leads to the total lock up behavior you report. If this is the case then moving the target location for the large copy activity to the PERC RAID group should help.

This should be a very localized problem (ie affect just this server) if you are on a healthy switched network but if the other servers have dependencies on shares or services hosted by that server then they too could exhibit some issues. That said the symptoms you describe hint at something more severe than that. If you physically log onto one of these servers during one of these incidents do you see the same problem?

1
  • Hi, If I log onto one of the other servers at the time the NAS hangs I just get a blank console after entering credentials. It will eventually respond and bring up the desktop. If I try to bring up the run box it will hang for a couple of seconds then respond. This doesn't happen to every server though. I'll run perfmon and try and capture the memory activity and post back. Many thanks for your comments so far. Jim.
    – user25272
    Nov 9, 2009 at 10:54
0

I've had better results with the server since matching the switch port to the NIC speed. I've also turned off the AMON module on NOD32 as I noticed that when the transfers were happening sometimes the the AV seemed to hang onto the file for a period of time. I'm running 2.7 on the server as I had huge problems with V4 on most of my servers.

The server is still not 100% but I don't seem to experience any long hangs anymore. Also the servers memory didn't seem to be the cause of the issue.

0

When the problem happens are you running into problems because logging into a second server will try to load a profile from the NAS server?

How large are the large files? Does it lock the server each and every time? Can you test a large file copy between two workstations and see it that works properly, with the AV running on both ( or a second server ).

I know someone had stated above to look at windows memory usage....

I would look at the network card drivers , you already said the drivers are up to date ,in that case I would look for an older driver, or turning off the tcp offloading options for the NIC. I had to work with MS for a windows cluster network issue, it is good to turn off nic offloading when troubleshooting.

HTH, Mark

You must log in to answer this question.