Red Hat Bugzilla – Bug 449139
RH5.2 Xen Windows GOS Network failure
Last modified: 2016-04-26 09:46:39 EDT
Description of problem:
My windows 2003 Guest OS loses network communication. My Xen server and Linux
guest OSes do not show these symptoms. On of the differenced between my
windows and linux GOS is that Windows is Fully Virtualized while linux is
running as a paravirtualized VM. Not sure how to get more information on this
issue and the windows event logs don't show much.
Version-Release number of selected component (if applicable):
Found on Redhat 5.2 snapshot 7
Happens sporadically from time to time (usually a couple of times a day)
Steps to Reproduce:
1. Start an application that sends network traffic to the windows VM. I also
had traffic being sent to my linux VM so that may have had somehting to do
Does this still happen with the GA version? Can you supply the following:
output of ipconfig /all from the guest and the dom0 dmesg output when this
Created attachment 309655 [details]
Image with the IP config information
Here is the requested ipconfig information after the issue has happened. I was
not able to capture the image right after the problem since it happened at 5 in
My guest OS is Windows 2003 R2 Enterprise Edition Service Pack 2 and I have
not seen the issue with a fully virtualized Linux Guest OS.
can you capture the dmesg output as suggested by bill?
does "Sending traffic" mean Pinging?
Then I tried with win2003 R2 SP2 and over a period of 3 hours, there is no packet loss and no network drop.
The Host is :2.6.18-128.1.1.el5xen #1 SMP Mon Jan 26 14:19:09 EST 2009 x86_64 x86_64 x86_64 GNU/Linux
By "Sending Traffic" what I meant was that the link be tested to ensure that it remains up. Our test application continuously monitors the VM through the network while generating IO to the storage on the VM. I would think that pinging would suffice as a test although it may not detect all cases.
Also your host version "2.6.18-128.1.1.el5xen" I believe is from a 5.3 install and I do not recall seeing this issue with a 5.3 host (although I had the para-virtualized drivers loaded). I saw this on a 5.2 xen server setup.
Well, I tried it and no luck to reproduce it. You wrote something about your test application - could you please send the test application for us to allow testing (just a binary, it's Windows system anyway :) Are you talking about application running in Windows guest, right ?).
Anyway, is that test application designed to send traffic all the time by itself with no need of user interaction with some logs being created? This could help a lot to have this application and try it by ourselves...
The testbench that we use is internal HP proprietary. Also, it requires a separate linux host and is not too strait forward. The primary focus of the application is to generate storage traffice, but it keep tabs on the servers/VMs to see how things are going and the such. This is the part that is failing. The test application starts doing its storage traffic thing then a few hours down the road, it loses communication with the VM.
Well, what do you mean by generating storage traffic? What sort of packets are sent there? Just for clarification, is this the application that is running inside the Windows VM? Also, what do you mean by few hours - how long it has been tested since this happened? Five hours? Eight hours? More?
Basically its a bunch of reads and writes going down to the storage device (Virtual disks presented to the VM). The main application runs on the linux server that I mentioned above but there are client processes that are started on the Windows VM itself. The main process then checks on the client processes through the network, but there isn't a lot of network traffice generated. Its mostly status messages get sent back and forth. The test usually fails after about 5 hours.
Could you please try to reproduce it with Jirka's RPMs from:
and tell if this is the issue?
Hi Hector, any luck with this one? Did you try this with RPMs from comment #19 ? You could also try with: http://people.redhat.com/jdenemar/xen/ and could you please provide results what's working and what's now working (which RPMs)? I was unable to reproduce it at all.
since I am unable to reproduce it, could you please try with RPMs from http://people.redhat.com/minovotn/xen and provide us the test results? Also, could you give us the exact steps to reproduce and what your Windows application in the guest is exactly doing ?
I guess no testing was done with RPMs described in c#22 so I need to ask you more about your application used for testing. You wrote it's basically a bunch of read/write operations going down to the VM's disks with main application running on Linux server. Should I consider this communication as client/server communication where client resides in the guest (VM) and server is standard application for Linux system? What's the communication protocol? Or this doesn't matter and it's just sending random data from guest (VM) to host machine (server)? Just random data ? You wrote something about checking the data by main process on the client processes on the Windows VM itself - by main process you mean the Linux server application or what? By client process you mean the one inside Windows VM, ie. this is Windows application, right ?
Since reporter is no longer available to help with debugging assistance and since we were unable to reproduce it at all with RHEL 5.4 version I am going to close it. If you run into those issues please feel free to reopen this one with more information.
This bug was closed during 5.5 development and it's being removed from the internal tracking bugs (which are now for 5.6).