Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

Looking for some help with a problem that has me at my wit's end - I've troubleshot this for days and cannot figure it out.

All of a sudden, a few days ago, copying files from some of our Windows XP SP3 x86 workstations to one particular Windows 2008 R1 x64 server has become super-slow - think 7 minutes to transfer a 10MB file. The problem is only workstation -> server, copying in the other direction is working normally.

Copying files to this same server (and same file share) has been fine for months previously, and as far as I know nothing has been changed on the server, workstations, Group Policy, etc. The workstations are physical machines, the server is a VM running in ESX 3.5, everything is connected by gigabit LAN and all are joined to the same (Windows 2008 funtioncal level) domain.

There is nothing obviously wrong on either the workstations or the server - no CPU/memory/disk issues or spikes, no event log entries, no apparent DNS or Active Directory issues, etc. Also, apart from this specific problem, the workstations and the server are behaving entirely normally (including network copy to other servers / shares).

Through some trouble-shooting I've established that problem is only occurring on some of our workstations - specifically, three machines that are used by our IT department. This does mean slightly different Group Policy and application set, but as I mentioned above, nothing should have changed around when the problem started, and none of these machines have anything unusual installed on them that should affect the network or file sharing.

Another unusual aspect to this problem is that it's occurred once before - involving exactly the same workstations and a different server, but in both cases the problem servers are almost identical - Windows 2008 x64 VM, running IIS7 as their /only/ app, being used as our development web server. Last time round we just nuked the server (and replaced it with the one that has the issue this time) which fixed the problem until now, but given the problem has repeated itself I want to get to the root of it.

Here's what I've tried so far, all to no avail:

  • Rebooted :-)
  • Disabled anti-virus and firewall.
  • Turned off every service possible on the server.
  • Re-installed VMware Tools on the server.
  • Updated network drivers on the workstations.
  • Used different user accounts - it's machine specific, not user specific.
  • Created new shared folders / shares on the server.
  • Used several different copying methods - Explorer, TeraCopy and xcopy.
  • Mapped the share using IP, NetBIOS name and FQDN.
  • Flushed DNS and ARP caches.
  • Forced DNS re-registration.
  • Fiddled with Network Card properties (link speed, Flow Control, TOE & TSO options, MTU, etc).
  • Uninstalled IIS7 on the server (thinking this was the common denominator between the two servers we've had issues with).
  • Probably some other stuff I've forgotten by now...

I also tried capturing a network trace with WireShark. I don't know much about analysing these, but I did compare the trace of a "normal" copy to the trace of a "super-slow" copy, and the main difference seems to be lots of fairly long pauses (typically ~0.3 seconds) before a series of error entries beginning with things like "[TCP Retransmission]", "[TCP Dup ACK...", "[TCP Fast Retransmission]" and "[TCP Out-Of-Order]". Not sure if that's any help.

So - anyone have any bright ideas? I'm at a loss as to what could be wrong or how to fix it :-S

share|improve this question
    
You haven't said that you have checked the switch for errors on the port, I would try that next. –  Richard Slater Feb 3 '10 at 8:46
    
Can't see anything happening at the switch port - surely if it was a hardware / switch error it wouldn't just affect three workstations to one server though (especially given the server is a VM, so its physical network card is shared by lots of other VMs too), and you also wouldn't expect the problem to start at exactly the same time on three different workstations? –  Matt Feb 4 '10 at 21:35
    
Boy I hope you find an answer, I have an ALMOST identical problem, except it's 2008R2, and running under VMWare server; and SMB is the only thing affected, I can FTP to it at practically wire speed. I've done a lot of the same fiddling with no luck. –  SqlACID Feb 5 '10 at 1:03
    
A switch problem can affect more than one port - depends on the switch and how much shared hardware is behind each port. Also if you have multiple uplinks on an ESX vSwitch with one iffy connection then some clients connecting to some VM's might have problems, and not others (if you are using ip based hashing for load balancing). Definitely worth swapping the ESX server uplinks to known good ports if you can do it without downtime. And check the cables too. –  Helvick Feb 9 '10 at 14:52

2 Answers 2

I would check the windowing on both servers. Server 2003 and XP have newer IP stacks and sometimes they fight aginst each other.

The retransmits and dupe acks are one server saying "slow down" or doing a source quench to make the other server feed data slower.

Try this site for some better explanation: link text

share|improve this answer
up vote 0 down vote accepted

Well, we've managed to resolve this problem - except I'm not exactly sure how :-)

The problem has completely gone away, both on the server I posted about above, and also on the older development server I mentioned where we'd had this problem previously.

The only things that have changed on / affecting these servers are:

  • Upgraded our hypervisor from ESX v3.5u4 to vSphere v4.0u1
  • Upgraded the VMware Tools inside the VMs from the ESX v3.5u4 version to the vSPhere v4.0u1 version.
  • Replaced the virtual network cards inside the VMs (we used these VMs as testbeds for the new VMXNET3 and Para-virtualised SCSI drivers in vSphere 4, which involved removing the old virtual NICs and adding new virtual NICs).

So, while I don't know what exactly fixed it, it's highly likely to be one of the above three items - hopefully that gives anyone else experiencing the problem some ideas to try.

Thanks for the responses all.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.