Anyone successfully hosting a Windows Terminal Server on Proxmox? I've been experiencing intermittent crashes on a 64-bit Windows 2003 Terminal Server (25 users), and I'm starting to think that perhaps this should just be a standalone physical server. It works absolutely fine at times and then it just crashes. Microsoft attributes the problem to "APIC latency". Any thoughts or experiences would be appreciated.
Windows Terminal Services
- Thread starter mazer9
- Start date
I've recently installed proxmox 1.7 on a dual xeon cpu server, one guest win2003-32 and the other -64bit, both R2-SP2. I've been told by a coworker that tonight he was transferring files from a GNU/Linux server to Win2003-64 through WinSCP and the system "stopped" abruptly (the guest switched off) twice, the first time after lots of GB (18), the second time after much less. I use virtio for hd and network, proxmox server has 1gbit and old GNU/Linux 100Mbs, GB switch in between.
Is the first time we have a dual processor server, and first time we virtualize (and use) win2003-64 (R2-SP2)
He was working through the VNC interface of Proxmox, tunneled through ssh.
I'm very scared, don't know if related to the "APIC latency", just hoping KVM people to find the problem in a short time and proxmox to update kvm with the patch.
Let's keep in touch about this problem
Best regards
Is the first time we have a dual processor server, and first time we virtualize (and use) win2003-64 (R2-SP2)
He was working through the VNC interface of Proxmox, tunneled through ssh.
I'm very scared, don't know if related to the "APIC latency", just hoping KVM people to find the problem in a short time and proxmox to update kvm with the patch.
Let's keep in touch about this problem
Best regards
# pveversion -v
pve-manager: 1.7-10 (pve-manager/1.7/5323)
running kernel: 2.6.32-4-pve
proxmox-ve-2.6.32: 1.7-30
pve-kernel-2.6.32-4-pve: 2.6.32-30
qemu-server: 1.1-25
pve-firmware: 1.0-10
libpve-storage-perl: 1.0-16
vncterm: 0.9-2
vzctl: 3.0.24-1pve4
vzdump: 1.2-10
vzprocps: 2.0.11-1dso2
vzquota: 3.0.11-1
pve-qemu-kvm: 0.13.0-3
ksm-control-daemon: 1.0-4
proxmox:~#
I hope to be able to update qemu-server soon to 1.1-28 (there has been a problem in proxmox report about this recently solved). Virtio drivers are red hat 1.1.11-0
pve-manager: 1.7-10 (pve-manager/1.7/5323)
running kernel: 2.6.32-4-pve
proxmox-ve-2.6.32: 1.7-30
pve-kernel-2.6.32-4-pve: 2.6.32-30
qemu-server: 1.1-25
pve-firmware: 1.0-10
libpve-storage-perl: 1.0-16
vncterm: 0.9-2
vzctl: 3.0.24-1pve4
vzdump: 1.2-10
vzprocps: 2.0.11-1dso2
vzquota: 3.0.11-1
pve-qemu-kvm: 0.13.0-3
ksm-control-daemon: 1.0-4
proxmox:~#
I hope to be able to update qemu-server soon to 1.1-28 (there has been a problem in proxmox report about this recently solved). Virtio drivers are red hat 1.1.11-0
That´s what I will try today. At first I thought it is 2.6.35-related (vhost-net). I´m using virtio driver 1.1.16 and the regtweaks posted on linux-kvm.com. My problems are disconnects at high guest network-load. I hope the performance is not decreasing too much with e1000.
No, because with "out of the box" e1000 driver there are big problems:
http://forum.proxmox.com/threads/41...ctly-for-win-2003-R2-an-d-kernel-2.6.32-2-pve
and I've not been able to find proper Win2003 R2 drivers on Intel site (sigh!).
Now we have another anomaly as well. If you log with vnc and download a file from internet with the browse, everything is ok. If you log in RDP, as soon as the "save as" dialog box opens the connectivity is stopped (does not respond to ping anymore). You have to login in vnc, disable and re-enable the nick to have it work again. This, my co-worker says, happens with a rate of 80% of the time he tries it.
At proxmox level, vmbr0 seems ok all the time. Only anomaly I've noticed is a:
vmtab101i0d0 Link (101 is the vm with such problems)
with
TX packets:285589562 errors:0 dropped:0 overruns:508 carrier:0
(some overruns).
Btw, I've applied the suggested Win regitry changes for have maximum throughput (see proxmox wiki), maybe this could be the problem? Anyone of the many Virtio settings that can solve this issue?
Thanks for the attention
http://forum.proxmox.com/threads/41...ctly-for-win-2003-R2-an-d-kernel-2.6.32-2-pve
and I've not been able to find proper Win2003 R2 drivers on Intel site (sigh!).
Now we have another anomaly as well. If you log with vnc and download a file from internet with the browse, everything is ok. If you log in RDP, as soon as the "save as" dialog box opens the connectivity is stopped (does not respond to ping anymore). You have to login in vnc, disable and re-enable the nick to have it work again. This, my co-worker says, happens with a rate of 80% of the time he tries it.
At proxmox level, vmbr0 seems ok all the time. Only anomaly I've noticed is a:
vmtab101i0d0 Link (101 is the vm with such problems)
with
TX packets:285589562 errors:0 dropped:0 overruns:508 carrier:0
(some overruns).
Btw, I've applied the suggested Win regitry changes for have maximum throughput (see proxmox wiki), maybe this could be the problem? Anyone of the many Virtio settings that can solve this issue?
Thanks for the attention
proxmox1:~# pveversion-v
-bash: pveversion-v: command not found
proxmox1:~# pveversion -v
pve-manager: 1.7-10 (pve-manager/1.7/5323)
running kernel: 2.6.32-4-pve
proxmox-ve-2.6.32: 1.7-28
pve-kernel-2.6.32-4-pve: 2.6.32-28
qemu-server: 1.1-25
pve-firmware: 1.0-9
libpve-storage-perl: 1.0-16
vncterm: 0.9-2
vzctl: 3.0.24-1pve4
vzdump: 1.2-9
vzprocps: 2.0.11-1dso2
vzquota: 3.0.11-1
pve-qemu-kvm: 0.13.0-2
ksm-control-daemon: 1.0-4
On a KVM forum, I've read about other people having problems running terminal services on KVM with intermittent crashes:
http://forums.meulie.net/viewtopic.php?f=43&t=5095
Additionally, I've read where VMware recommends only one vCPU for a Terminal Server or Citrix Server due to CPU contention:
http://communities.vmware.com/message/792304
Whereas, I've yet to discover any best practices documentation for Terminal Services on KVM (or any regarding successful implementation for that matter), I wonder if the same issues for VMware based VMs apply here?
I'll do some testing and report back...
-bash: pveversion-v: command not found
proxmox1:~# pveversion -v
pve-manager: 1.7-10 (pve-manager/1.7/5323)
running kernel: 2.6.32-4-pve
proxmox-ve-2.6.32: 1.7-28
pve-kernel-2.6.32-4-pve: 2.6.32-28
qemu-server: 1.1-25
pve-firmware: 1.0-9
libpve-storage-perl: 1.0-16
vncterm: 0.9-2
vzctl: 3.0.24-1pve4
vzdump: 1.2-9
vzprocps: 2.0.11-1dso2
vzquota: 3.0.11-1
pve-qemu-kvm: 0.13.0-2
ksm-control-daemon: 1.0-4
On a KVM forum, I've read about other people having problems running terminal services on KVM with intermittent crashes:
http://forums.meulie.net/viewtopic.php?f=43&t=5095
Additionally, I've read where VMware recommends only one vCPU for a Terminal Server or Citrix Server due to CPU contention:
http://communities.vmware.com/message/792304
Whereas, I've yet to discover any best practices documentation for Terminal Services on KVM (or any regarding successful implementation for that matter), I wonder if the same issues for VMware based VMs apply here?
I'll do some testing and report back...
I've had the same problem of Win2k8 + TS + Xenapp randomly crashing..with BSOD 0x0000000a
I've made tests with IDE and VIRTIO disks, e1000 with fresh Intel drivers, Virtio nic, 1 or 2 cores : always the same troubles..
I didn't notice anything in kernel.log
I had no other choice than running it on a physical server...
do you have the same BSOD code?
I've made tests with IDE and VIRTIO disks, e1000 with fresh Intel drivers, Virtio nic, 1 or 2 cores : always the same troubles..
I didn't notice anything in kernel.log
I had no other choice than running it on a physical server...
do you have the same BSOD code?
I received mostly x3B (System_Service_Exception) and x7E (System_Thread_Exception) crashes.
Check here for detailed information related to the crashes:
http://forum.proxmox.com/threads/5317-Windows-2003-64-Bit-Random-Crashes-APIC-related
Perhaps a standalone physical server is the only remedy for this problem. Or so it seems based on the feedback.
Check here for detailed information related to the crashes:
http://forum.proxmox.com/threads/5317-Windows-2003-64-Bit-Random-Crashes-APIC-related
Perhaps a standalone physical server is the only remedy for this problem. Or so it seems based on the feedback.
That is the reason I want to test a 1000Hz proxmox kernel. But I couldn´t get it working. Please can someone help me to remedy the issue? http://forum.proxmox.com/threads/5328-PVE-Kernel-compile?p=30285#poststop
No like I said I want to test it but the compilation is not working as expected (http://tinyurl.com/2u3bv8q) and I really compiled many kernels on my ubuntu box. I think some git-commands reverting my settings to default. Or should I try a vanilla-kernel with your patches?
Thank you very much
Thank you very much
I did some further experiment on my own. The problem fired up as soon as the "save as" dialog appears, but with IE and not Firefox. Tshark showed lots of
[TCP Retransmission] [TCP segment of a reassembled PDU] and DUP ACK errors around that moment.
I then downloaded and installed latest e1000 drivers from Intel site (version 8.10.3.0) and disabled virtio nic and enabled e1000 one, rebooted.
Everything seems to work fine now! (hope will do also tomorrow) I'm stunned
Hope the other problem (sudden crash during transfer) will never happen too.
Still wondering if is just a matter of fine-tune the many nic parameters of Virtio, or just it's broken and no one has yet be able to provide useful bug report / fix the issue.
Thanks a lot of your attention and help
[TCP Retransmission] [TCP segment of a reassembled PDU] and DUP ACK errors around that moment.
I then downloaded and installed latest e1000 drivers from Intel site (version 8.10.3.0) and disabled virtio nic and enabled e1000 one, rebooted.
Everything seems to work fine now! (hope will do also tomorrow) I'm stunned
Hope the other problem (sudden crash during transfer) will never happen too.
Still wondering if is just a matter of fine-tune the many nic parameters of Virtio, or just it's broken and no one has yet be able to provide useful bug report / fix the issue.
Thanks a lot of your attention and help
Ok I tried the e1000 now and it is stable but slow now. what about the 1000Hz howto from my last post?
I get this speed with my actual e1000-config (Windows 2003 32bit, Intel drivers updated)
That is slow compared to my virtio-config (948 Mbits/sec) but it is stable now, no disconnects via RDP and SMB.
Perhaps I found the problem with my host-config. I mixed two types of NIC´s (2x Intel and 2x Broadcom) in one 802.3ad bond. These NIC´s have different offloading capabilities. So I will test again with bonding only one type of NIC and doing the virtio-tests again.
Happy new year to all the Proxmox-Fans
macday
That is slow compared to my virtio-config (948 Mbits/sec) but it is stable now, no disconnects via RDP and SMB.
Perhaps I found the problem with my host-config. I mixed two types of NIC´s (2x Intel and 2x Broadcom) in one 802.3ad bond. These NIC´s have different offloading capabilities. So I will test again with bonding only one type of NIC and doing the virtio-tests again.
Happy new year to all the Proxmox-Fans
macday