Hi yaniv,
Thanks for your response. My answers below.
cordialement, regards,
Nelson LAMEIRAS
Ingénieur Systèmes et Réseaux / Systems and Networks engineer
Tel: +33 5 32 09 09 70
nelson.lamei...@lyra-network.com
www.lyra-network.com | www.payzen.eu
Lyra Network, 109 rue de l'innovation, 31670 Labège, FRANCE
From: "Yaniv Kaul" <yk...@redhat.com>
To: "Nelson Lameiras" <nelson.lamei...@lyra-network.com>
Cc: "ovirt users" <users@ovirt.org>
Sent: Wednesday, April 19, 2017 9:13:27 AM
Subject: Re: [ovirt-users] massive simultaneous vms migrations ?
On Tue, Apr 18, 2017 at 7:48 PM, Nelson Lameiras <
nelson.lamei...@lyra-network.com > wrote:
hello,
When putting a host on "maintenance mode", all vms start migrating to other
hosts.
We have some hosts that have 60 vms. So this will create a 60 vms migrating
simultaneously.
Some vms are under so much heavy loads that migration fails often (our guess is
that massive simultaneous migrations does not help migration convergence) -
even with "suspend workload if needed" migraton policy.
- Does oVirt really launches 60 simultaneous migrations or is there a queuing
system ?
- If there is a queuing system, is there a way to configure a maximum number of
simultaneous migrations ?
I did see a "migration bandwidth limit", but this is quite what we are looking
for.
What migration policy are you using?
-- "suspend workkiad if needed"
Are you using a dedicated migration network, or the ovirtmgmt network?
-- ovirtmgmt network
BQ_BEGIN
my setup:
ovirt-engine +hosted engine 4.1.1
hosts : centos 7.3 fully updated.
for full context to understand this question : 2 times in the past, when trying
to put a host in maintenance, host stopped responding during massive migrations
and was fenced by engine. It's still unclear why host stopped responding, but
we think that migrating 60+ vms simultaneously puts a heavy strain on storage ?
So we would like to better control migration process in order to better
understand what's happening. This scenario is "production only" since our labs
do not contain nearly as much vm with such heavy loads. So rather than trying
to reproduce, we are trying to avoid ;)
BQ_END
If you could open a bug with relevant logs on the host not responding, that'd
be great.
-- Too late, we made the "precipitaded" mistake of reinstalling server (beause
of other reasons), so all traces are lost. Next time I will make sure to keep
traces.
Live migration doesn't touch the storage (disks re not moving anywhere), but it
does stress the network. I doubt it, but perhaps you over-saturate the
ovirtmgmt network.
-- This makes sens. It's a long shot though. Maybe it would be possible to
create a VLAN dedicated to migrations in same physical network et use QOS to
reserve always some bandwith to oVirt administration ? (maybe opening can of
worms here....)
Y.
BQ_BEGIN
cordialement, regards,
Nelson LAMEIRAS
Ingénieur Systèmes et Réseaux / Systems and Networks engineer
Tel: +33 5 32 09 09 70
nelson.lamei...@lyra-network.com
www.lyra-network.com | www.payzen.eu
Lyra Network, 109 rue de l'innovation, 31670 Labège, FRANCE
_______________________________________________
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users
BQ_END
_______________________________________________
Users mailing list
Users@ovirt.org
http://lists.ovirt.org/mailman/listinfo/users
