Jorgen Lundman
2010-Apr-26 08:21 UTC
Console output blocks domU when console buffer gets full.
dom0: SunOS xen01.unix 5.11 snv_134 i86pc i386 i86xpv Compiled against library: libvir 0.7.0 Using library: libvir 0.7.0 Using API: Xen 3.0.1 Running hypervisor: Xen 3.4 domU: SunOS ssh00.unix 5.10 Generic_141445-09 i86pc i386 i86pc If I run a command that outputs a few lines on the console, ie, so that they show up when connected with: # virsh console ssh00.unix the system works great. If I disconnect "virsh console" (ctrl-]) and keep running the command on domU, they will continue to work until the console buffer is full. About ~24 lines of text or so. Then the domU will stop responding completely. If I reconnect to console, via "virsh console", the buffer is emptied, and the domU comes back to life. Is this a known problem? Normally, domUs probably very rarely send IO on console so that the buffer do not fill up, but with kernel module debugging, this was very noticeable. I suppose I could tell syslog.conf to not use console as well. -- This message posted from opensolaris.org
Jorgen Lundman
2010-Apr-27 00:42 UTC
Re: Console output blocks domU when console buffer gets full.
If you want to test it yourself, I run the following command on domU ''ssh00'' # while [ 1 ]; do logger -p kern.alert `date`; date; sleep 1; done Which shows output on domU ''ssh'' login as well as on dom0''s ''virsh console'' dom0: Apr 27 09:36:02 ssh00.unix lundman: Tue Apr 27 09:36:02 JST 2010 Apr 27 09:36:03 ssh00.unix lundman: Tue Apr 27 09:36:03 JST 2010 domU: Tue Apr 27 09:36:02 JST 2010 Tue Apr 27 09:36:03 JST 2010 I disconnect "virsh console": Apr 27 09:36:04 ssh00.unix lundman: Tue Apr 27 09:36:04 JST 2010 xen01:~# and domU keeps running: Tue Apr 27 09:36:05 JST 2010 Tue Apr 27 09:36:06 JST 2010 Tue Apr 27 09:36:07 JST 2010 [...] until: Tue Apr 27 09:36:58 JST 2010 Tue Apr 27 09:36:59 JST 2010 Tue Apr 27 09:37:00 JST 2010 and it is dead. You can not even ping the server. We connect back on "virsh console": xen01:~# virsh console ssh00.unix Apr 27 09:36:05 ssh00.unix lundman: Tue Apr 27 09:36:05 JST 2010 Apr 27 09:36:06 ssh00.unix lundman: Tue Apr 27 09:36:06 JST 2010 [...] Apr 27 09:37:00 ssh00.unix lundman: Tue Apr 27 09:37:00 JST 2010 Apr 27 09:37:27 ssh00.unix lundman: Tue Apr 27 09:37:27 JST 2010 (Bit of a pause there, 27 seconds gone?) and domU also resumes: Tue Apr 27 09:37:27 JST 2010 ^C -- This message posted from opensolaris.org
Mark Johnson
2010-Apr-27 11:35 UTC
Re: Console output blocks domU when console buffer gets full.
On 04/26/10 08:42 PM, Jorgen Lundman wrote:> If you want to test it yourself, I run the following command on domU ''ssh00''This is fixed in b137 by 6931744 DomU occasionally stops responding after reboot http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6931744 MRJ> # while [ 1 ]; do logger -p kern.alert `date`; date; sleep 1; done > > Which shows output on domU ''ssh'' login as well as on dom0''s ''virsh console'' > > dom0: > Apr 27 09:36:02 ssh00.unix lundman: Tue Apr 27 09:36:02 JST 2010 > Apr 27 09:36:03 ssh00.unix lundman: Tue Apr 27 09:36:03 JST 2010 > domU: > Tue Apr 27 09:36:02 JST 2010 > Tue Apr 27 09:36:03 JST 2010 > > I disconnect "virsh console": > > Apr 27 09:36:04 ssh00.unix lundman: Tue Apr 27 09:36:04 JST 2010 > xen01:~# > > and domU keeps running: > > Tue Apr 27 09:36:05 JST 2010 > Tue Apr 27 09:36:06 JST 2010 > Tue Apr 27 09:36:07 JST 2010 > [...] > > until: > > Tue Apr 27 09:36:58 JST 2010 > Tue Apr 27 09:36:59 JST 2010 > Tue Apr 27 09:37:00 JST 2010 > > and it is dead. You can not even ping the server. > > We connect back on "virsh console": > > xen01:~# virsh console ssh00.unix > Apr 27 09:36:05 ssh00.unix lundman: Tue Apr 27 09:36:05 JST 2010 > Apr 27 09:36:06 ssh00.unix lundman: Tue Apr 27 09:36:06 JST 2010 > [...] > Apr 27 09:37:00 ssh00.unix lundman: Tue Apr 27 09:37:00 JST 2010 > Apr 27 09:37:27 ssh00.unix lundman: Tue Apr 27 09:37:27 JST 2010 > > (Bit of a pause there, 27 seconds gone?) > > and domU also resumes: > > Tue Apr 27 09:37:27 JST 2010 > ^C