ZomboDB in PosgreSQL cluster mode and ZomboDB behavior with Partition/Sharding

[Frederic Brégier]

Hi,

I would like to know if there are any recomandations regarding using ZomboDB with a PostgreSQL Cluster ?

I imagine 2 scenarios:

1) First is an active/passive PostgreSQL cluster, classical one, using log shipping (through PostgreSQL configuration of course) between the primary server and the secondary server.

Is there any possibility to think of a mutualisation of the Elasticsearch index behind? Such that for instance

a) Primary is responsible to write to index

b) Primary and perhaps Secondary are able to read from the index

The idea is to keep one index for the full cluster, since Elasticsearch is already a cluster by itself and it seems unuseful to get twice the very same index.

Of course, when the Primary goes down, by a reconfiguration, the Secondary could take over the Primary role, including the write role to Elasticsearch index.

Do you think it is feasable ? and how ?

2) Now let's go to active/active mode, using for instance PG Pool II in front of the PostgreSQL cluster.

My intuition is that if we consider each server as a "standalone" one, there must be 2 indexes within Elasticsearch.

But considering the replication taking by PG Pool II, could it be possible to specify as in the previous scenario, 1 for Write and both for Read ?

In another way, for both scenarios : is that possible to specify using ZomboDB that one server is responsible to write Index or not, while still requesting in read being possible?

If not, of course, each and every servers would have to take its own index (so doubling the storage and Elasticsearch cluster while not really usefull).

Last but not least, what happens in case of Partition or even Sharding when using ZomboDB on Elasticsearch side ?

Thanks for your feedbacks

Frederic