Data360 DQ+

 View Only
  • 1.  Vertica DB

    Posted 03-24-2021 09:19

    I understand that the Ver0 ComputeDb Master is the primary Vertica DB instance. Does that mean that its content is replicated to all other Vertica DB nodes (In our Production Instance we Ver0 to Ver7) ? In case of Ver0 unavailability, will any other Node take over the Master role? If so the Replication happens how often ?



  • 2.  RE: Vertica DB

    Employee
    Posted 03-24-2021 14:16

    Hi Gopi,

    It looks like the vertica cluster isn't a high-availability cluster. it's a cluster for performance and Vertica requires 3 nodes at a minimum; it's just how the product is designed.  Vertica db can also be manually backed up via the following script residing in the the install bin directory that could potentially be configured in a cronjob to recurring backups too.  

    backupComputeDb
    If you want to backup and restore ComputeDB, you can follow the instructions in the guide "Restoring the system" found on page 56.
     
     


  • 3.  RE: Vertica DB

    Employee
    Posted 03-31-2021 05:36

    Hi Gopi,

    In addition to the information provided earlier, Vertica does not use a primary/standby concept. All of the nodes in the cluster are always active and participate in responding to queries. The replication between nodes is real time. Vertica's ability to tolerate node failures depends on the k-safety of the cluster. The default configuration of 3 nodes, has a k-safety of 1 which means that there is 1 additional copy of all the data on a node. This, in turn, means that the vertica cluster can survive 1 of the 3 nodes going down. However, due to some bugs being fixed at this time, if a non-master node fails, the application can handle it but it cannot handle a master node failure at this time without manual intervention. We are working on fixing this for a future release.  I hope it answers your questions.  Let us know if you still have any lingering questions.