Expand my Community achievements bar.

Learn about Edge Delivery Services in upcoming GEM session
SOLVED

Data Store configuration guidelines

Avatar

Level 5

Are there any guidelines on what kind of data store configuration (shared between publish instances, shared between author and publish, no sharing at all) is best for majority of the cases?

I've read in a no. of places that a store shared between author and publish may be the best choice because of the reduced storage requirements, reduced replication load, etc.

Is that the go to approach used these days?

1 Accepted Solution

Avatar

Correct answer by
Employee

Indeed a shared datastore is the best practice. It allows us to reduce disk space and simplify the replication by employing a binary less replication.

The configuration is not much diffident from a non-shared datastore. The min required arguments are minRecordLength and path.

Shared datastore requires some special DataStore Garbage Collection routine.  For a shared File Data Store, you need to run Datastore Garbage Collection with a mark only flag from all of AEM that are using a shared file data store. Next, you need to run Datastore Garbage Collection without mark only flag from one of AEM that is using a shared file data store.

View solution in original post

3 Replies

Avatar

Correct answer by
Employee

Indeed a shared datastore is the best practice. It allows us to reduce disk space and simplify the replication by employing a binary less replication.

The configuration is not much diffident from a non-shared datastore. The min required arguments are minRecordLength and path.

Shared datastore requires some special DataStore Garbage Collection routine.  For a shared File Data Store, you need to run Datastore Garbage Collection with a mark only flag from all of AEM that are using a shared file data store. Next, you need to run Datastore Garbage Collection without mark only flag from one of AEM that is using a shared file data store.

Avatar

Level 5
Thanks for the reply. With shared data store how do you mitigate the risks due to a single point of failure? That seems like a big drawback?