Hide Forgot
Description of problem: Customer reports that randomly his users are reporting that their spice/dual monitors are hanging and exiting. Version-Release number of selected component (if applicable): Host details: ------------- Hostname: dwi16.dwi.litle.com Distro: [redhat-release] Red Hat Enterprise Virtualization Hypervisor release 6.6 (20150603.0.el6ev) Booted kernel: 2.6.32-504.23.4.el6.x86_64 spice-server-0.12.4-11.el6.x86_64 vdsm-4.16.20-1.el6ev.x86_64 VM Details ---------- Hostname: vsarah-vm1 Distro: [redhat-release] Red Hat Enterprise Linux Server release 6.8 Booted kernel: 2.6.32-642.4.2.el6.x86_64 spice-client-0.8.2-15.el6.x86_64 rhevm-guest-agent-common-1.0.11-6.el6ev.noarch rhevm-guest-agent-gdm-plugin-1.0.11-6.el6ev.x86_64 How reproducible: Random however certain VMs see it more frequently. In this case vsarah-vm1 Steps to Reproduce: 1.User connects to RHEL 6.8 vm with dual monitors from a Windows laptop via spice. 2. After some time the monitors will hang and go white. 3. They must close the monitor windows and reconnect and even then spice options like cut and paste stop working. Actual results: Monitors hang Expected results: There should be no hangs Additional info: When this happens customer is generally running an application called squirrel and doing a query. We have verified that the VM was shutdown and properly set to allow to monitors before running with the dual monitors.
The resolution for the monitors is as follows: one screen is 1432x1087 the other screen is 1872x1106 These are not running in full screen mode. There are no indications in the logs as there being any hang or problem even with spice debugging enabled.
Hello, Please provide additional information: 1. What is the version of spice-vdagent package on the guest ? (on the guest -- rpm -q spice-vdagent) 2. What is OS is running on the laptop (e.g Windows 7 amd64 or Windows XP) ? What is the version of virt-viewer/remote-viewer running on the laptop ? (e.g. from Help->About menu) 3. Was the VM configured to have 2 monitors in RHEV-M ? (possibly in addition what was the qemu-kvm command line on the host) 4. Are there any errors/warnings in the qemu-kvm log file on the host ? (in /var/log/libvirt/qemu/VMNAME.log)
1 - rpm -qav|grep spice spice-gtk-python-0.26-7.el6.x86_64 spice-client-0.8.2-15.el6.x86_64 spice-glib-0.26-7.el6.x86_64 spice-gtk-0.26-7.el6.x86_64 spice-server-0.12.4-13.el6.1.x86_64 spice-protocol-0.12.6-1.el6.noarch spice-vdagent-0.14.0-11.el6.x86_64 spice-xpi-2.7-27.el6.x86_64 2 - windows 7 pro 64B virt viewer 3.1-256 3 - Yes I had them shutdown the VM and verify that they selected two monitors before starting the vm again. 4 - attached qemu log file no errors
Created attachment 1216140 [details] vsarah qemu log
Hi Frank, We also need: 1. log from the client to check what's going on there; You can run with --spice-debug to gather the logs 2. log from spice-vdagent in the guest based on: > and even then spice options like cut and paste stop working. I also suggest the customer to check with rhevm spice client (virt-viewer 3.1 is upstream release)
The customer has set the number of heads correctly as verified by the attached xml dump from the failing VM. I noticed that there is no entry for vgamem. This seems to be not implemented until rhev 3.6.3 with 7.2 hosts according to BZ 1275539. The white screens are always happening after multiple cut+paste operations. I suspect that this would only be solved by the fixed put into BZ 1275539. Does this sound correct? Thanks, Frank
Created attachment 1260399 [details] XML dump from failing VM
Logs from client and agent please, in order to see something unusual that could help create a reproducer (comment #5)
It seems we are seeing some clipboard activity and then the session is gone and they need to reconnect: Oct 4 11:16:24.082484 spice-vdagentd: debug: 0x60bed0 clipboard req from agent which is not in the active session? Oct 4 11:16:24.082562 spice-vdagentd: debug: 0x60c070 received clipboard grab, arg1: 0, arg2: 0, size 4 Oct 4 11:16:24.114917 spice-vdagentd: debug: 0x60bed0 received clipboard release, arg1: 0, arg2: 0, size 0 Oct 4 11:16:24.114967 spice-vdagentd: debug: 0x60bed0 clipboard req from agent which is not in the active session? Oct 4 11:16:24.114981 spice-vdagentd: debug: 0x60c070 received clipboard release, arg1: 0, arg2: 0, size 0 Oct 4 11:16:24.115744 spice-vdagentd: debug: 0x60bed0 received clipboard grab, arg1: 0, arg2: 0, size 4 Oct 4 11:16:24.115823 spice-vdagentd: debug: 0x60bed0 clipboard req from agent which is not in the active session? Oct 4 11:16:24.116385 spice-vdagentd: debug: 0x60c070 received clipboard grab, arg1: 0, arg2: 0, size 4 Oct 4 11:16:24.149587 spice-vdagentd: debug: 0x60bed0 received clipboard release, arg1: 0, arg2: 0, size 0 Oct 4 11:16:24.149643 spice-vdagentd: debug: 0x60bed0 clipboard req from agent which is not in the active session? Oct 4 11:18:21.810761 spice-vdagentd: debug: New client connected Oct 4 11:18:21.974484 spice-vdagentd: debug: 0x60c070 sent monitors config, arg1: 0, arg2: 0, size 48 Oct 4 11:18:21.976712 spice-vdagentd: debug: 0x60c070 sent monitors config, arg1: 0, arg2: 0, size 328 Oct 4 11:18:22.738557 spice-vdagentd: debug: 0x60bed0 received guest xorg resolution, arg1: 3840, arg2: 1200, size 32 Oct 4 11:18:22.738592 spice-vdagentd: debug: uinput-update-size: 3408x1134 Oct 4 11:18:22.738600 spice-vdagentd: debug: screen 0: +0+0 Oct 4 11:18:22.738612 spice-vdagentd: debug: screen 1: +1512+0 Oct 4 11:18:22.738620 spice-vdagentd: debug: 0x60bed0 received guest xorg resolution, arg1: 3840, arg2: 1200, size 32 Oct 4 11:18:22.738627 spice-vdagentd: debug: uinput-update-size: 3408x1134 Oct 4 11:18:22.738634 spice-vdagentd: debug: screen 0: +0+0 Oct 4 11:18:22.738641 spice-vdagentd: debug: screen 1: +1512+0 Oct 4 11:18:22.738648 spice-vdagentd: debug: 0x60bed0 received guest xorg resolution, arg1: 3840, arg2: 1200, size 32 Oct 4 11:18:22.738654 spice-vdagentd: debug: uinput-update-size: 3408x1134 Oct 4 11:18:22.738683 spice-vdagentd: debug: screen 0: +0+0 Oct 4 11:18:22.738690 spice-vdagentd: debug: screen 1: +1512+0 Oct 4 11:18:22.756210 spice-vdagentd: debug: 0x60c070 received guest xorg resolution, arg1: 3840, arg2: 1200, size 32 Oct 4 11:18:22.756267 spice-vdagentd: debug: uinput-update-size: 3840x1200 Oct 4 11:18:22.756301 spice-vdagentd: debug: screen 0: +0+0 Oct 4 11:18:22.756324 spice-vdagentd: debug: screen 1: +1920+0 Oct 4 11:18:22.782446 spice-vdagentd: debug: 0x60c070 received guest xorg resolution, arg1: 3840, arg2: 1200, size 32 Oct 4 11:18:22.782517 spice-vdagentd: debug: uinput-update-size: 3840x1200 Oct 4 11:18:22.782533 spice-vdagentd: debug: screen 0: +0+0 Oct 4 11:18:22.782547 spice-vdagentd: debug: screen 1: +1920+0 Oct 4 11:18:23.156981 spice-vdagentd: debug: 0x60c070 sent monitors config, arg1: 0, arg2: 0, size 328 From the customer the time of the failure: Thursday, October 04, 2018 11:18 AM I am attaching the complete spice-vdagent log but it is HUGE.
Thanks Frank, this is very important hint > spice-vdagentd: debug: 0x60bed0 clipboard req from agent which is not in the active session? As the messages between spice components could be doing what is expected but something is broken in the agent leading to weird behaviour. Any other information from the RHEL 6? Is it running GNOME? Was it tested with latest RHEL 6 version too? I'm asking because lasted build for spice-vdagentd is 0.14-13 while comment #3 mentions usage of 0.14-11
They are running 6.10 in the guest with the latest spice-vdagent. kernel-2.6.32-754.3.5.el6.x86_64 spice-vdagent-0.14.0-13.el6.x86_64 It does appear that they are using gnome within the virtual machine. gnome-desktop-2.28.2-11.el6.x86_64
Reproduced the warnings so far: > Jan 8 11:01:03.040009 spice-vdagentd: debug: 0xfe5510 received clipboard grab, arg1: 1, arg2: 0, size 4 > Jan 8 11:01:03.040322 spice-vdagentd: debug: 0xfe91b0 received clipboard grab, arg1: 1, arg2: 0, size 4 > Jan 8 11:01:03.040510 spice-vdagentd: debug: 0xfe91b0 clipboard req from agent which is not in the active session? 0xfe5510 and 0xfe91b0 are both spice-vdagent from gdm user; Having two instances of spice-vdagent for the same user is the bug and it is related to console-kit. To trigger this scenario, I have here a RHEL-6.10 VM image from QE: 1) user 'test' connects automatically. At this point, only 'test' session agent is running; 2) Switch user will open the GDM interface; Connect with 'test2' user 3) At this point you have 3 session agents running: 'gdm', 'test', 'test2' 4) Switch user will open the GDM interface; Log back to 'test' user; 5) At this point, you have 4 session agents running: 'gdm', 'gdm', 'test', 'test2'
> Having two instances of spice-vdagent for the same user is the bug and it is related to console-kit. Err. I think we can somehow workaround using console-kit/logind but the issue is related to whom is creating the multiple daemons (...) Reproducible also in a RHEL7.6 VM $ ps aux | grep -i spice | grep -v "root\|grep" gdm 3805 0.0 0.1 190172 3936 ? Ss 14:42 0:00 /usr/bin/spice-vdagent test 4452 0.0 0.1 190172 3892 ? Ss 14:42 0:00 /usr/bin/spice-vdagent -d gdm 5120 0.0 0.1 190172 3932 ? Ss 14:43 0:00 /usr/bin/spice-vdagent toso 5742 0.0 0.1 190172 3892 ? Ss 14:43 0:00 /usr/bin/spice-vdagent -d