Re: Postgresql BDR(Bi-Directional Replication) Latency Monitoring

[Craig Ringer <craig@xxxxxxxxxxxxxxx>]

Hi

I just replied to a fairly similar question here:

http://www.postgresql.org/message-id/CAMsr+YEqeFXDS=AmqBZMopQDFnt+hO50uV-+s854KW9oA-XKkA@xxxxxxxxxxxxxx

and I'll follow up on that thread as it's essentially the same thing.

On 17 March 2015 at 02:27, Bill Brown <billb@xxxxxxxxxxxxxxxxxxx> wrote:

Hi All,

I'm looking at  BDR monitoring documentation at this location https://wiki.postgresql.org/wiki/BDR_Monitoring

I understand the query in documentation:

select slot_name, plugin, database, active, xmin,

            pg_get_transaction_committime(xmin)

            FROM pg_stat_logical_decoding ;

Should be:

select slot_name, plugin, database, active, xmin,

            pg_get_transaction_committime(xmin)

            FROM pg_replication_slots ;

I am running PostgreSQL 9.4.0 bits on Centos:

postgresql-bdr94-contrib-9.4.0_bdr1-1_2ndQuadrant.el6.x86_64

postgresql-bdr94-libs-9.4.0_bdr1-1_2ndQuadrant.el6.x86_64

postgresql-bdr94-bdr-0.8.0beta1-1_2ndQuadrant.el6.x86_64

postgresql-bdr94-2ndquadrant-redhat-1.0-2.noarch

postgresql-bdr94-devel-9.4.0_bdr1-1_2ndQuadrant.el6.x86_64

postgresql-bdr94-server-9.4.0_bdr1-1_2ndQuadrant.el6.x86_64

postgresql-bdr94-9.4.0_bdr1-1_2ndQuadrant.el6.x86_64

 

We have 3 nodes set-up.  We would like to understand the latency between nodes.

When using http://oltpbenchmark.com/ to create a load on one node,  I would expect the 'xmin' value not to be null but is.  

What is the best way to monitor replication latency when using Postgresql BDR?

Regards,

Bill

--

 Craig Ringer                   http://www.2ndQuadrant.com/
 PostgreSQL Development, 24x7 Support, Training & Services