Does AEM 6 still support clustering (not mongo)

[B_Wulteputte]

When looking at the documentation, it looks regular TarMK clustering has disappeared and clustering is now only supported for mongoDB.

So the core question is here, does the following still apply http://docs.adobe.com/docs/en/cq/current/deploying/cluster.html for AEM 6 and if so, do the steps still match? The documentation seems to suggest clustering is only supported for mongoDB (http://docs.adobe.com/docs/en/aem/6-0/deploy/configuring/cluster.html)

[Sham_HC · Accepted solution]

[1] will be valid for aem6 if you are using crx2.

Not valid for crx3 or oak, For crx3 you need to go with mongo. Other option might be to install hotfix Hot fix 5354  and use standby[2] option.

[1]   http://docs.adobe.com/docs/en/cq/current/deploying/cluster.html

[2]   http://docs.adobe.com/content/docs/en/aem/6-0/deploy/recommended-deploys/tarmk-cold-standby.html

[artie_leland]

Sham, you say "For crx3 you need to go with mongo." 

Could you elaborate on why you think CRX3 clustering will not work with TarMK? I am looking for some explanation that is more than RTFM :)

[Sham_HC]

Artie Leland wrote...

Sham, you say "For crx3 you need to go with mongo." 

Could you elaborate on why you think CRX3 clustering will not work with TarMK? I am looking for some explanation that is more than RTFM :)

  What is your case for crx3 cluster ? so that i will try to answer accordingly.

[autowoven]

Ok, here's some more background:

We are currently in production using AEM 5.6. We have 6 publish instances, each with it's own repository. A single Author instance replicates content to these 6 Publish instances.

As part of the upgrade to AEM 6.0, we would like to move the repository to Amazon S3 - for both scalability and performance. At the same time, we would like to optimize the replication process. We would like the Author instance to publish to a single Publish instance that uses TarMK microkernel with the repository in S3. The other 5 Publish instances should just read the content from the S3 repository and therefore not need replication.

Our ideal goal is to use TarMK (we do not want to use Mongo DB) on CRX3, and would only like to replicate to a single instance. We do want multiple publish instances (that talk to this single repository) for network (and other) performance reasons.

[justin_at_adobe]

Hi,

A few points to clarify:

* TarMK never supported clustering. TarMK is new technology in Oak. I think you are confusing it (for good reason) with TarPM.

* S3 cannot be used to host the nodestore, only the datastore. So you still need to use TarMK or MongoMK for the nodestore.

* It is not clear to me why you think that using S3 will improve performance. In general, S3 will be at par with or slower than local disk. The S3 datastore is really designed for large scale storage. Of course, your mileage may vary based on the specifics of your application.

* In order to avoid replication of binaries, you must use the *same* datastore in author and publish and then set up binaryless replication. You can do this with TarPM, TarMK, or MongoMK. 

Regards,

Justin