[OOTB-infra] Virtualisation platforms - IRC discussion

[Martin Cosgrave]

Nov 3rd, 11:39 - 12:05 CET

<marsbard> hi-ko: I'd just like to be very sure, before we commit to a 
proprietary product, that we can't get along with a FOSS product, I 
think it would be better if the bee stuff was FOSS if possible. ohej, 
bmejias, do you guys have an opinion on this?
<marsbard> hi-ko: presumably we could still transfer vms but it would 
need an extra step of converting them to ovf I suppose
<marsbard> I mean to take them to vmware if we had to
--> twen (~twen at unaffiliated/twen) has joined #orderofthebee
<hi-ko> the vm will not contain any closed license stuff at all just 
open-vm-tools - but will us make sure it runs in the customer 
environment wich is in most cases not kvm. So it's just a question of 
the runtime and how to avoid/solve problems as efficient as possible
<bmejias> marsbard, I haven't follow the discussion, but in terms of 
"principles", I agree that I'd prefer to have as much as possible with 
FOSS software
<marsbard> who are the customers? from ootb PoV?
<bmejias> marsbard, but I'm not religious in that sense. If something is 
correctly justified, AND there is no FOSS alternative...
<marsbard> bmejias: I'm totally in agreement there, but the kvm 
solutions look really good, judging from my short time looking at them 
over the weekend
<marsbard> of course if hi-ko is going to do the infrastructure install 
then it will of course be fastest to build on what he knows
<hi-ko> marsbard, bmejias: I don't agree with runtime. We did the error 
in the past to choose for a FOSS virt platform and nobody cares about 
that huge amount of extra work you have to to if you don't run on best 
practice platforms
<hi-ko> marsbard: OOTB end customers? --> has to be discussed ;-)
<marsbard> hi-ko: indeed, to be discussed, so we can hardly make 
judgements about what they need, can we? ;-)
<marsbard> or can I presume you have some ideas you are thinking about? ;-)
<bmejias> in any case, we use VMWare over here hosting plenty of RHEL's, 
and I don't have any say on that.
<bmejias> on my previous job we were using KVM for the local instances...
<bmejias> so, I don't have a strong opinion....
<bmejias> marsbard's argument regarding "who is doing the job" and that 
"it is better that is done in something he knows about" is a good one
<marsbard> I think it is easiest to go with vmware now if hi-ko is going 
to do the install for us (I hope I'm not being presumptuous about that 
hi-ko!)
<marsbard> personally I am going to try to get some hardware with 
working virtualisation capabilities and play with some of the other options
<marsbard> actually my buddy who has a server at the same place has a 
kvm setup I believe so maybe I can try out the tools on his server
<hi-ko> marsbard: I don't have the time this moment to discuss in detail 
but we have ~8 years histroy in set up and maintain alfresco systems in 
end customer environments. I haven't seen KVM once. We supported Xen 
somewhen in the past but it was too much work for 2-3 customers in 
allover europe
<hi-ko> I'm not docmatic - just pragmatic
<bmejias> marsbard, yeah, until we haven't someone grokking the FOSS 
alternative, we better go for the "know-how" choice
<marsbard> ok, thanks hi-ko and bmejias
<hi-ko> marsbard: May the force bee with you
<marsbard> then again... there is enough budget to cover two servers? 
could it be worth getting a second and using it to prove the case for 
alfresco vms on KVM? So we can get moving quickly on our core servers 
via ESXi and then also use the FOSS tools for other things
-*- marsbard looks over the conversation to see if anything needs to be 
summarised in the mailing list
<hi-ko> two platforms for which added value?
<hi-ko> no fail over - no interchange ...
<hi-ko> marsbard: idea is the backup one server from the other which 
only makes sense if they run the same virtualisation
<marsbard> I didn't know we were planning on failover
<hi-ko> kvm is the best choise if the knowledge and support is available 
on maintainers side
<hi-ko> not real failover, but hardware will fail some day and it's just 
vm start on the other side to be online again
<marsbard> I'm not sure that we are such a wealthy organisation that we 
want to have a server just sitting there waiting for the other one to 
fail. I would rather have another working server and if the hardware 
fails we take the downtime and restore from backups
<hi-ko> marsbard: no the server just saves the backup from the other in 
case something fails one can host both
--> douglascrp (~douglascr at pluto.konsultex.com.br) has joined #orderofthebee
<marsbard> I see what you're saying hi-ko, e.g. 3 VMs on each, and if 
one host goes down the other hosts 6 VMs temporarily
<hi-ko> marsbard: exactly - or you decide to temporarily stop vm 4 or 
what ever
<marsbard> ok
<hi-ko> server 1 runs all systems which are visible from outside and 
should perform well, server 2 runs all test systems and may slow down if 
a build process or a benchmark is running