Re: Two instances of Postgres with single data directory

John R Pierce <pierce@xxxxxxxxxxxx> · Fri, 14 Nov 2014 10:45:52 -0800



    On 11/14/2014 2:24 AM, Albe Laurenz
      wrote:

    
      dineshkaarthick wrote:

      
        > I would like to know how is the "Shared Disk Failover" replication achieved
> if it is not possible to share the data directory ? I am referring to the
> 1st solution in the mentioned link,
> http://www.postgresql.org/docs/9.3/static/different-replication-solutions.html

      
      You have some cluster software that makes dead sure that one node is *not*
running the database server before starting it on the other.
    
    
    more specifically, that the data volume isn't even MOUNTED by the
    other server.   typically this is enforced with hardware 'fencing'
    such that the standby server is blocked from even trying to mount
    the shared volume, until it decides to take over, then it fences the
    original 'master', unfences itself, and mounts the volume, then
    starts its copy of postgresql.    this fencing is often done in a
    storage switch (fiberchannel most commonly).

    
    -- 
john r pierce                                      37N 122W
somewhere on the middle of the left coast