No problem, glad I could help.

Regards,
Dag Sonstebo
Cloud Architect
ShapeBlue

On 21/11/2016, 19:19, "Grégoire Lamodière" <g.lamodi...@dimsi.fr> wrote:

    Hi Dag, 
    
    Thank you for this feedback.
    
    I will check for a smooth migration (migration is not a big deal, but 
working with IP reservation, etc. could be time consumming).
    Anyway, I think my understanding was not the good one regarding pr-storage, 
and situation is now very clear thanks to your last reply about the network 
labels.
    
    Best Regards.
    
    ---
    Grégoire Lamodière
    T/ + 33 6 76 27 03 31
    
    -----Message d'origine-----
    De : Dag Sonstebo [mailto:dag.sonst...@shapeblue.com] 
    Envoyé : lundi 21 novembre 2016 18:31
    À : users@cloudstack.apache.org
    Objet : Re: Zone edit - Physical network update
    
    Hi Grégoire,
    
    Obviously hoping to work with you, but if you were a support customer we 
would advise against this approach – moving all CloudStack traffic to a new 
physical network is very much untested territory and we couldn’t guarantee 
success. We would probably advise you to consider a phased swing migration – 
where you potentially free up some of your existing hypervisors, rebuild these 
with the correct networking and then add them as a new cluster in CloudStack – 
before trying migrate your workload. As you then free up more resources on the 
original cluster you can evacuate more hosts, rebuild them and add to the new 
cluster, until you have all your hosts in the new cluster. 
    
    With regards to you “cs-mgmt” traffic question  – keep in mind there is 
nothing stopping you just keeping this where it is on eth0. You don’t need to 
use labels for *primary storage*  – this simply requires connectivity and 
CloudStack doesn’t care how this works. The CloudStack storage network 
labelling is purely for secondary storage. So if you add the new cards as 
effectively IP based HBAs with an additional storage IP address and present new 
primary storage pools purely over this interface you can simply add new NFS 
shares as new primary storage pools. This means the XenServers will speak to 
old primary storage on the existing interfaces (cs-mgmt?) and to the new 
primary storage over the new interfaces (call this e.g. cs-primarystorage if 
you want). This would potentially allow you to storage migrate from old to new 
primary storage *without changing any existing CloudStack labelling*.
    
    If you require more advanced traffic configuration changes than this the 
phased migration mentioned above would be your best bet.
    
    Regards,
    Dag Sonstebo
    Cloud Architect
    ShapeBlue
    
    On 21/11/2016, 16:47, "Grégoire Lamodière" <g.lamodi...@dimsi.fr> wrote:
    
        Hi Dag, 
        
        Thanks for your quick reply.
        
        As you can easily understand, the goal of this lab job is to be moved 
to production later on, so un-predictable options are not really options 
(specially when you'll know that Shapeblue will probably handle the support for 
us in the coming weeks:))
        
        The goal is to move our current SAS 6gbits storage to 10 Gbe, so yes, I 
was assuming management network to a new network card (10 Gbe), and storage as 
well (to take advantage of 10 Gbe rather than staying on the 1 Gbe link).
        
        When you say " Keep in mind the physical network is mainly cosmetic", 
does it stand for :
        
                "Physical Network 1
                        Management traffic (ex. : cs-mgmt label)"
        
        Doesn't mean that cs-mgmt network should be related to eth0 on the 
physical host ? 
        In other work, can I plug my new network card, rename cs-mgmt on the 
old one to cs-mgmt-old and create cs-mgmt on the new one ?
        
        If so, you are right, I will add the card and move the management label 
to the 10 Gbe.
        But I assume I could do the same for the storage network label ?
        
        Thanks again.
        
        Best Regards.
        
        ---
        Grégoire Lamodière
        
        -----Message d'origine-----
        De : Dag Sonstebo [mailto:dag.sonst...@shapeblue.com] 
        Envoyé : lundi 21 novembre 2016 16:59
        À : users@cloudstack.apache.org
        Objet : Re: Zone edit - Physical network update
        
        Hi Gregoire,
        
        First of all – I think you’re probably on very thin ice with this one – 
you may be able to do this in the lab but it would be very risky to carry out 
in a production environment.
        
        Saying that – if you want to persevere and are happy this may break 
your lab - a starting point would possibly be:
        
        - Add the new cards to your XenServers.
        - Create new networks in XenCentre – with new network labels.
        - Not sure if this is required – but you would probably add the new 
physical network using CloudMonkey. Keep in mind the physical network is mainly 
cosmetic – your network configuration uses labelling only hence you may just 
use the existing physical network.
        - Stop CloudStack management.
        - Hack the network labels in the cloud.host_details and 
cloud.physical_network_traffic_types to reflect the new networks.
        - On your XenServers / Xencentre move all existing networks (VLANs) to 
the new networks (unsure of the steps or if this is possible).
        - <insert other magic here>
        - Restart management.
        
        One other point – you mention this is for a *primary storage* migration 
– keep in mind the storage network in the CS configuration is for *secondary 
storage*. If all you want to do is to migrate to a new storage network and a 
new storage backend your best bet is to:
        
        - Add the new cards + configure the new storage network on the 
XenServers – no need to label this.
        - Present new primary storage pools from your NAS on the respective 
networks – making sure the new NFS shares are only available over your 10Gbe 
network.
        - Add the new NFS shares as new primary storage pools in the ACS gui.
        - Storage migrate from old NFS shares to new.
        
        This is a much simpler way and would be a supported configuration 
change.
        
        Hope this helps,
        
        Regards,
        Dag Sonstebo
        Cloud Architect
        ShapeBlue
        
        
        On 21/11/2016, 14:50, "Grégoire Lamodière" <g.lamodi...@dimsi.fr> wrote:
        
            Dear All,
            
            In my lab, I have a CS 4.7 advanced zone working fine.
            
            I am now in the process of testing a primary storage migration 
(currently SAS based to 10 Gbe nfs).
            To perform this migration, I need to attach new network controlers 
(10 Gbe controlers)
            
            I have no problem for the hardware part of my work, but I am 
wondering how CS can handle this change.
            
            Right now, on my zone conf, I have :
            
            
            -          Physical Network 1
            
            o   Management traffic (with XenServer Label)
            
            -          Physical Network 2
            
            o   Public traffic
            
            o   Guest
            
            o   Storage
            
            What is the best « CS way » to add my card (each with 2 10 Gbits 
ports), add move management + Storage on the first port of the new card ?
            
            -          Physical Network 1
            
            o   No more traffic
            
            -          Physical Network 2
            
            o   Public
            
            o   Guest
            
            -          Physical Network 3 (first port of the new 10 Gbe card)
            
            o   Management
            
            o   Storage
            
            
            I tried to play arround with create physical network, but I can't 
do much with the newly created network (no way to assign traffic type to them).
            
            Any help would be highly appreciated.
            
            Best Regards.
            
            ---
            Grégoire Lamodière
            
        
        
        dag.sonst...@shapeblue.com
        www.shapeblue.com
        53 Chandos Place, Covent Garden, London  WC2N 4HSUK @shapeblue
          
         
        
        
    
    
    dag.sonst...@shapeblue.com
    www.shapeblue.com
    53 Chandos Place, Covent Garden, London  WC2N 4HSUK @shapeblue
      
     
    
    


dag.sonst...@shapeblue.com 
www.shapeblue.com
53 Chandos Place, Covent Garden, London  WC2N 4HSUK
@shapeblue
  
 

Reply via email to