git.net

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Upgrading OS on hypervisors


No magic as Ivan said,

you may want to check dynamic auto-convergence in case of busy VMs (high
RAM change ram i.e. busy MSSQL servers and such) - qemu 2.5 needed for
this, can be consumed from 4.11 and onwards - but might not be relevant in
your case.

Some very short interruption IS done - i.e. source VM on source host is
paused - and last few bits/megabits of RAM are synced to new VM on new host
- then source destination VM is killed/destroyed and destination VM (which
was paused all time...) will be un-paused/resumed - this should have be
less than a second or so...

No magic to make it happen without a brief interruption to networking - the
faster management network the better (better throughput, so less pause
period for VM...)


Cheers

On Fri, 14 Dec 2018 at 07:45, Rakesh v <www.rakeshv.com@xxxxxxxxx> wrote:

> Hello Ivan
>
> We upgrade only when critical security patch is released like meltdown and
> L1TF else we don't do upgrades for every single kernel release
>
> Sent from my iPhone
>
> > On 13-Dec-2018, at 11:05 PM, Ivan Kudryavtsev <kudryavtsev_ia@xxxxxxxxx>
> wrote:
> >
> > Rakesh,
> > No magic exists. Migrate forth, reboot, and migrate back. What I'm really
> > wondering about is why you upgrade every time, Ubuntu releases a new
> > kernel... There are a pretty small amount of fixes related to KVM and
> hosts
> > are located in the private networks, they are single-tenant. Maybe you
> just
> > have to change to policy for that?
> >
> > чт, 13 дек. 2018 г. в 16:54, Rakesh v <www.rakeshv.com@xxxxxxxxx>:
> >
> >> Hello Folks
> >>
> >> I have a question regarding upgrading OS on hypervisor. We have few
> >> platforms with around 200 hypervisors in each platforms. Every time a
> new
> >> Ubuntu kernel comes in with new security patch like L1TF fix or other ,
> we
> >> need to upgrade kernel on all hypervisors. Before rebooting hypervisor
> we
> >> enable maintenance on it so that all vm's are migrated away. During this
> >> maintenance we always see some Network interruptions because of live
> >> migrations.
> >>
> >> How are you guys handling such situations and how are doing the OS
> >> upgrades? If you have some suggestions or using some automation tools
> then
> >> please let know
> >>
> >> I want to reduce time taken to upgrade all hypervisors without any
> impact
> >> to Networks or VM's
> >>
> >> Sent from my iPhone
> >
> >
> >
> > --
> > With best regards, Ivan Kudryavtsev
> > Bitworks LLC
> > Cell RU: +7-923-414-1515
> > Cell USA: +1-201-257-1512
> > WWW: http://bitworks.software/ <http://bw-sw.com/>
>


-- 

Andrija Panić