Hello Sarah.
I guess you are right, your explanation makes a lot of sense but
to confirm that I'll need to wait for a while to give a
maintenance window and make new tests.
The server will be under a heavy production for the next days.
I'll return soon.
Thank you for your answer.
Andrey
On 06-09-2016 19:57, Sarah Newman wrote:
On 09/06/2016 06:29 AM, Andrey Schmiegelow wrote:
Hello everybody.
After a lot of experiences with others kernels and it's parameters I found a
workaround.
Removing the parameter "console=XXX " from the kernel line a new PV guest
deploys (beyond the others 50 PV guests already running with this parameter).
Of course, after that I can't access this new guests through xl console command.
Until now this problem seems to happen only with PV guests while I have no
problems deploying more HVM guests.
This workaround if fine for my current purpose, but it will be interesting
understand why this happen.
I will back if I have any news.
Thank you for all you help me.
I believe that adding extra consoles will fail if the maximum number of open files available to xenconsoled is too low. There might be a distribution
specific way to increase it. You can look in /proc/<pid>/limits for the current value. That number can be adjusted dynamically with prlimit with a new
enough kernel, or you can increase the allowed amount in /etc/security/limits.conf and restart xenconsoled.
To my best knowledge, the reason you're seeing a problem with PV and not HVM domains is that (assuming you're not using stubdomains) qemu opens a pts
directly rather than going through xenconsoled.
--Sarah
_______________________________________________
Xen-users mailing list
Xen-users@xxxxxxxxxxxxx
https://lists.xen.org/xen-users
