Hide Forgot
Created attachment 817992 [details] engine log Description of problem: VM ticket was not set upon migration so the virt-viewer showed just password prompt. This error happened to me just once around 17:26 local time (16:26 UTC). The VM name is rhel6-64-33-dj and VM UUID is dc636bfb-9fd2-45f4-874b-9e997ebf0d29 Version-Release number of selected component (if applicable): is20.1 - engine and dst host: vdsm-4.13.0-0.5.beta1.el6ev.x86_64 libvirt-0.10.2-29.el6.x86_64 qemu-kvm-rhev-0.12.1.2-2.414.el6.x86_64 spice-server-0.12.4-6.el6.x86_64 source host is is19: vdsm-4.13.0-0.3.beta1.el6ev.x86_64 libvirt-0.10.2-29.el6.x86_64 qemu-kvm-rhev-0.12.1.2-2.414.el6.x86_64 spice-server-0.12.4-6.el6.x86_64 How reproducible: just once Steps to Reproduce: 1. do back and forth migrations with client connected 2. 3. Actual results: client asked for password, qemu log says that the ticketing is enabled but password is not set (this is correct spice behaviour) Expected results: password is set Additional info: component is just a blind guess, the issue may be anywhere in the stack...
martin, please check in the logs, david, since it happened only once and under not really normal circumstances - moving out to 3.4
The destination logs indicate that the qemu failed to hand over the ticket: ((null):9709): Spice-Warning **: reds.c:2719:reds_handle_read_link_done: spice channels 1 should be encrypted ((null):9709): Spice-Warning **: reds.c:2820:reds_handle_ssl_accept: SSL_accept failed, error=5 red_dispatcher_loadvm_commands: ((null):9709): Spice-Warning **: reds.c:1966:reds_handle_ticket: Ticketing is enabled, but no password is set. please set a ticket first
I'm not sure if anything else can be added without knowing what spice-server was thinking…
It does not look like a spice-server problem. If the ticket was not set, spice-server can not accept connections. The dst_libvirt_log, which is a suspect, does not contain information about the problematic VM. The first attempt of the client to connect, just before migration started, failed (SSL_accept failed with error 5 == SSL_ERROR_SYSCALL). The second attempt to connect (after migration completed) failed, because no ticket was set.
David, have you hit that issue just once, or is it still occurring for you ?
I didn't use RHEV for a while so I didn't hit it recently but it just popped up time to time. It was neither frequent, nor one-time issue. Have you had a look to code paths that Uri pointed at?
Client logs could probably help there :(
closing for now, feel free to reopen with more details on how to reproduce & logs thanks