locked
Cluster or spare server? RRS feed

  • Question

  • Hi all,

     

    I'm currently running on a single HOST server with a SAN to store VHD and DATA of virtualized file server.

    The budget was not plan for this type of architecture, but buying a standard file server didn't make sense to me as 2-3 servers (physical) have to be change this fiscal year.

    So for next fiscal year (May), I have ask for a small budget (regarding 3 servers) to have the same host server as now and virtualized the 3 servers to be replaced.

     

    Now, my poor redudancy situation (only one host, that's a problem even with 4hours support), would become a nice situation.

     

    But my new concern is how to use this new host ?

    - Setting up a cluster?

    - Having a second host server, and manage "disaster" manualy?

    I know that cluster would give us a perfect availability, but it sounds very difficult to setup, and I can read so many problems here that I'm a little bit afraid to go in this direction.

    Our need are not like production companies, for example, the most important server have a 24h delay to come back in limited access, 36h back as usual!
    (that's a long time regarding computers, but it was like that (time to setup a new physical server, put backup to it) in the past)

    The second solution sounds very easy for me. I will be able to setup VM manually. Dispatch the charge (where to run the VM), and turn on VM's from Host1 on Host2 manually in case of issue with Host1 hardware.

    As VHDs are stored on a LUN, they will be available even if one host is "dead".

    I'm already thinking of 2 LUNs to avoid any issue with simultaneous access to LUN.
    I will simply mount the required LUN in case of emergency.

    What is your advice / experience.

    I'm looking for Cluster info, maybe positiv feedback that could put me on this direction.
    What are the real plus of this solution (if you don't speak about automatically switching)? 
    Is there any advalue? Like "load ballancing" I mean moving VM from host1 to host2 to dispatch memory / CPU use?

     

    Regards, Yannick

     

     

     

     

     

     

     

     

     

     

    Tuesday, March 15, 2011 4:15 PM

Answers

  • Yes, there are some benefits of clustering your potential 2 servers together.  The question is, is it worth it?  There is some complexity in setting up a server and some more complexity in upgrading them, but over all Hyper-V clusters are getting better.  If  your organization needed very high uptime or the applications are critical then I would say, you need to cluster your servers, but as you mentioned, "the most important server have a 24h delay to come back in limited access, 36h back as usual",  you have a situation where a cold spare might be benefical.  I have many Hyper-V clusters in my environment, but also have branch offices that have Hyper-V standalone servers. 

    In my branch office senario, where there is greater tollerance for downtime, we have exact hardware for the Hyper-V hosts.  We have distributed cold spare hardware to certain sites so that a spare is no more than 1 hour away from a different site.  This allows us to keep our costs down so that we do not have a wasted spare at each site.  The most common problems we have had in the offices are blown power supplies, memory, and motherboard issues.  In these cases the cold spare is setup, disks and HBAs(if applicable) are swapped and the spare is powered back on.

    This is not a glamorous solution, but is, like many, a solution that takes technology, budget, and downtime tollerance into consideration.

     

    Rob McShinsky (http://www.VirtuallyAware.com)


    http://www.VirtuallyAware.com
    • Proposed as answer by Vincent Hu Wednesday, March 16, 2011 3:02 AM
    • Marked as answer by Yannick CLEVY Wednesday, March 16, 2011 8:42 AM
    Tuesday, March 15, 2011 5:13 PM
  • Yannick, personally I would prefer a Failover Cluster solution. You can have an active-active cluster, or an active-passive cluster. Generally speaking, mostly runs active-active cluster, and are able to load balance. 

    See some guidance for failover clustering with Hyper-V: http://kristiannese.blogspot.com/2011/01/failover-cluster-for-use-with-hyper-v.html

    I would also suggest you to enable CSV (Clustered Shared Volumes) on your LUNs, to take advantage of live migration.

    When it comes to administration and management overhead of the cluster, you really have to look at the benefits to be able to weigh that against each other.
    It gives you redundance, easier maintenance on your physical nodes in the cluster, simplified administration of your VMs, and much more.   In addition, you can also create guest clusters within the VMs, for your SQL server etc. The key word here is ‘highly available’.

    As you may know, Failover Clustering requires an Active Directory domain. It is recommended to place this role outside your cluster. I have written some posts about this topic on my blog, check it out here:  http://kristiannese.blogspot.com/2010/11/failover-clustering-and-domain.html and here: http://kristiannese.blogspot.com/2010/11/failover-cluster-and-domain.html



    Kristian (Virtualization and some coffee: http://kristiannese.blogspot.com )
    • Proposed as answer by Vincent Hu Wednesday, March 16, 2011 3:02 AM
    • Marked as answer by Yannick CLEVY Wednesday, March 16, 2011 8:42 AM
    Tuesday, March 15, 2011 5:13 PM
  • Hi

    As the other guys have said deciding on whether your setup warrants a cluster is your first hurdle. If you have 36h to get a server back online then I would say a cold spare would be an adequate solution.

    There are a lot of benefits to clustering but they do take some setting up and its striking the balance between your 'business case' needs and the desire to try clustering.

    IT at my work is critical hence a 5 node failover cluster and then I went and complicated it by clustering SQL within the Hyper-V cluster.

    The whole process was recorded on my blog www.edutechnow.com If you do manage to take a look and have any questions then just shout.


    Alan Richards www.edutechnow.com
    • Marked as answer by Yannick CLEVY Wednesday, March 16, 2011 8:42 AM
    Tuesday, March 15, 2011 6:50 PM

All replies

  • Yannick, personally I would prefer a Failover Cluster solution. You can have an active-active cluster, or an active-passive cluster. Generally speaking, mostly runs active-active cluster, and are able to load balance. 

    See some guidance for failover clustering with Hyper-V: http://kristiannese.blogspot.com/2011/01/failover-cluster-for-use-with-hyper-v.html

    I would also suggest you to enable CSV (Clustered Shared Volumes) on your LUNs, to take advantage of live migration.

    When it comes to administration and management overhead of the cluster, you really have to look at the benefits to be able to weigh that against each other.
    It gives you redundance, easier maintenance on your physical nodes in the cluster, simplified administration of your VMs, and much more.   In addition, you can also create guest clusters within the VMs, for your SQL server etc. The key word here is ‘highly available’.

    As you may know, Failover Clustering requires an Active Directory domain. It is recommended to place this role outside your cluster. I have written some posts about this topic on my blog, check it out here:  http://kristiannese.blogspot.com/2010/11/failover-clustering-and-domain.html and here: http://kristiannese.blogspot.com/2010/11/failover-cluster-and-domain.html



    Kristian (Virtualization and some coffee: http://kristiannese.blogspot.com )
    • Proposed as answer by Vincent Hu Wednesday, March 16, 2011 3:02 AM
    • Marked as answer by Yannick CLEVY Wednesday, March 16, 2011 8:42 AM
    Tuesday, March 15, 2011 5:13 PM
  • Yes, there are some benefits of clustering your potential 2 servers together.  The question is, is it worth it?  There is some complexity in setting up a server and some more complexity in upgrading them, but over all Hyper-V clusters are getting better.  If  your organization needed very high uptime or the applications are critical then I would say, you need to cluster your servers, but as you mentioned, "the most important server have a 24h delay to come back in limited access, 36h back as usual",  you have a situation where a cold spare might be benefical.  I have many Hyper-V clusters in my environment, but also have branch offices that have Hyper-V standalone servers. 

    In my branch office senario, where there is greater tollerance for downtime, we have exact hardware for the Hyper-V hosts.  We have distributed cold spare hardware to certain sites so that a spare is no more than 1 hour away from a different site.  This allows us to keep our costs down so that we do not have a wasted spare at each site.  The most common problems we have had in the offices are blown power supplies, memory, and motherboard issues.  In these cases the cold spare is setup, disks and HBAs(if applicable) are swapped and the spare is powered back on.

    This is not a glamorous solution, but is, like many, a solution that takes technology, budget, and downtime tollerance into consideration.

     

    Rob McShinsky (http://www.VirtuallyAware.com)


    http://www.VirtuallyAware.com
    • Proposed as answer by Vincent Hu Wednesday, March 16, 2011 3:02 AM
    • Marked as answer by Yannick CLEVY Wednesday, March 16, 2011 8:42 AM
    Tuesday, March 15, 2011 5:13 PM
  • Hi

    As the other guys have said deciding on whether your setup warrants a cluster is your first hurdle. If you have 36h to get a server back online then I would say a cold spare would be an adequate solution.

    There are a lot of benefits to clustering but they do take some setting up and its striking the balance between your 'business case' needs and the desire to try clustering.

    IT at my work is critical hence a 5 node failover cluster and then I went and complicated it by clustering SQL within the Hyper-V cluster.

    The whole process was recorded on my blog www.edutechnow.com If you do manage to take a look and have any questions then just shout.


    Alan Richards www.edutechnow.com
    • Marked as answer by Yannick CLEVY Wednesday, March 16, 2011 8:42 AM
    Tuesday, March 15, 2011 6:50 PM
  • Thanks to you guys and to Vincent Hu who avoid a copy paste of the first comment.

     

    I will study your links, but as I can see, in my case, it's 50/50.

    I'm afraid about complexity of cluster for a none desired high availability.

    I don't want to go on a "cold spare", but use it !

    Just having the possibility to move VMs from a broken HOST to the working one, even if performance will get lower during repair of the HOST.

    SO that's not money given for nothing (or just security), that's already use.

    Will come back in the next month, to share my experience whatever is the solution.

     

    Regards, yannick.

     

    Wednesday, March 16, 2011 8:46 AM