Forum Discussion

cmoreno1978's avatar
cmoreno1978
Level 3
14 years ago

Global Cluster Option with two single node cluster?

Hi all,

 

The question is simple, i`m checking the requeriments for this configuration, Cab we configure a global cluster between two single node cluster?

According to the characteristics of the customer platform, the only configuration possible to us is to implement a global cluster that is responsible for monitoring replication groups and fail over applications in case of serious failure in the main site.

I understand that we can configure a single-node cluster but I need  help  to consider any request or cons if we set a global cluster with two single node cluster

 

Thanks in advance,

business continuity
Cluster Server
  • mikebounds's avatar
    mikebounds
    14 years ago

    I am guessing you mean you need "Automated" failover rather than automatic failover.  What I mean is that you should set Global service groups to have Manual failover and if node dies, then you will get an alert to failover.  On accepting the alert, VCS will "automate" the failover.  I would not recommend setting Global service groups to have Auto failover as if at the time of failover you are in async mode (which you could be in even with soft sync), VCS will take over even if VVR is behind.  Also any network issues between the 2 sites could cause split-brain.  With an RTO of 4 hours, automatic failover should not be required.

    Whether to go async or soft sync is down to:

    1. Bandwidth - you can use vradvisor to see if your application ever exceeds 1Mbps - this is small, so I am guessing it might
    2. Latency - it can be quite hard to guage to tell what the effect the latency will have on your application, so you can just test soft sync and see how your application handles this - you can switch between async and soft sync on the fly, so it is not a "hard" setting

    If your RPO is 1 hour, then async should suffice and if your bandwidth is sufficient, then in you should only be a second or 2 behind.

     

    Mike.


  • Gaurav_S's avatar
    Gaurav_S
    Moderator
    14 years ago

    Hello,

    Yes, that is very much possible...

     

    From VCS 5.1 admin guide for solaris located at:

     

    https://sfdoccentral.symantec.com/Storage_Foundation_HA_51_Solaris.html

     

     

    Prerequisites for global clusters


    This section describes the prerequisites for configuring global clusters.

    Cluster setup


    You must have at least two clusters to set up a global cluster. Every cluster must
    have the required licenses. A cluster can be part of one global cluster. VCS
    supports a maximum of four clusters participating in a global cluster.
    Clusters must be running on the same platform; the operating system versions
    can be different. Clusters must be using the same VCS version.
    Cluster names must be unique within each global cluster; system and resource
    names need not be unique across clusters. Service group names need not be
    unique across clusters; however, global service groups must have identical
    names.
    Every cluster must have a valid virtual IP address, which is tied to the cluster.
    Define this IP address in the cluster’s ClusterAddress attribute. This address is
    normally configured as part of the initial VCS installation. The IP address must
    have a DNS entry.
    The global cluster operations require that the port for Wide-Area Connector
    (WAC) process (default is 14155) are open across firewalls.
    All clusters in a global cluster must use either IPv4 or IPv6 addresses. VCS does
    not support configuring clusters that use different Internet Protocol versions in
    a global cluster.
    For remote cluster operations, you must configure a VCS user with the same
    name and privileges in each cluster.

     

     

    Gaurav

  • mikebounds's avatar
    mikebounds
    Level 6
    14 years ago

    You can configure a global cluster with two single node clusters, but you should consider RDC (replicated Data cluster) if your nodes are fairly closed together (less than 50km, but possibly further apart).  You also need to consider what you are trying to provide - most people trying to configure this kind of set-up are trying to provide both HA and DR and this doesn't really work.  A few points on GCO v RDC:

    1. RDC is designed for sites closer together and GCO is designed for sites far apart.
       
    2. Failover is automatic by default for RDC and manual for GCO (both can be configured for Automatic or Manual failover)
       
    3. Detection of node failure (so that a failover is invoked) is by default 15 seconds for RDC and 3 – 4 minutes for GCO (both can be tuned for different values, however an RDC heartbeats continuously and GCO only heartbeats periodically, so GCO cannot be tuned to have as low node failure detection times as RDC without compromising the risk of split-brain.

    If you want auto failover, then you need to think carefully about split-brain and also the case where replication is async at the time of failover.

    So a few questions:

    1. How far apart are your nodes
    2. Are you trying to provide HA or DR (or both)
    3. Are you using VVR or Hardware replication
    4. What mode of replication are you using - hard sync (fail writes if network is down), soft sync (replicate sync, but switch to async if network is down), or async
    5. What is your RPO and RTO
    6. Do you require automatic failover.

     

    Mike

  • cmoreno1978's avatar
    cmoreno1978
    Level 3
    14 years ago

    Thanks you guys for quick response and help

     

    To answer the question:

     

    1. More than 100 Km

    2. DR, through the WAN with a bandwidth of 1 mbps approx. We have clarified to the customer that for HA and Local failover they need for additional hardware

    3. VVR

    4. Async or maybe Soft sync, wich are the best option considering the bandwith and type of application? ( is a WEB Server that it refers to a database of not more than 50 GB)

    5. RTO= 4 hrs

        RPO= 1 hour

    6. Yes.

     

    Thanks again and regards,

  • mikebounds's avatar
    mikebounds
    Level 6
    14 years ago

    I am guessing you mean you need "Automated" failover rather than automatic failover.  What I mean is that you should set Global service groups to have Manual failover and if node dies, then you will get an alert to failover.  On accepting the alert, VCS will "automate" the failover.  I would not recommend setting Global service groups to have Auto failover as if at the time of failover you are in async mode (which you could be in even with soft sync), VCS will take over even if VVR is behind.  Also any network issues between the 2 sites could cause split-brain.  With an RTO of 4 hours, automatic failover should not be required.

    Whether to go async or soft sync is down to:

    1. Bandwidth - you can use vradvisor to see if your application ever exceeds 1Mbps - this is small, so I am guessing it might
    2. Latency - it can be quite hard to guage to tell what the effect the latency will have on your application, so you can just test soft sync and see how your application handles this - you can switch between async and soft sync on the fly, so it is not a "hard" setting

    If your RPO is 1 hour, then async should suffice and if your bandwidth is sufficient, then in you should only be a second or 2 behind.

     

    Mike.