Binary less replication is not working

[vikask76191633]

I am having shared datastore between my local author and publisher, there is a default publish replication agenet with binaryless settings and url. Its for AEM6.3.

My Scenario:

Created and Authored Page with assets present at /content/dam/project and publish, it ask to publish the assets as well, If I select assets then dam assets along with page will be replicated to publisher, so what is the meaning of being binaryless, or Am I doing something fundamentally wrong ?

Regards

Vikas

[Gaurav-Behl]

In any case, only the metadata should be replicated and not the binary file. Could you please share the replication logs?

[Jörg_Hoh]

To actually get the binaryless replication working, you need to have a shared datastore with all instances (in your case: author & publish) using it.

[venkateshs73486]

Hi Jorg,

One question. My client wants to have Standalone external datastore at author &  publisher . ( not shared datastore).

Reason is, Client does not want publisher and site to fail in the event of shared datastore crashing..

But, wants to have provision to replicate the metadata of Image / Video assets.

But , for Non-image / Non-video assets, client wants to have normal binary-full replication.

Does this use-case need custom replication agent/ transport handler to resolve?

[Jörg_Hoh]

The standard replication can deal with both binaryless replication and "standard replication". But "standard replication" will always replicate the binaries. I guess that this is not the thing you want, correct?

But what do you want to do when you only have the metadata of an asset, but not the asset itself?

[venkateshs73486]

Thanks Jorg.

Client has DM-Scene7 setup. So, live page assets are actually pulled from Scene7 ..not from Publisher.

Metadata replication is to get all Scene7 properties intact and for tracking purpose.

However, assets like PDF need to be replicated with binaries to Publisher ( as Scene7 would not support hosting them)

[Jörg_Hoh]

In that case you can use the standard replication, but just configure an "Agent User" to each replication agent. This agent should have a wildcard ACL which prevents it from seeing the binary nodes.

This is probably going to work from an replication point of view, but I haven't tested what happens on a publish, if such an asset without binaries is getting replicated.

[venkateshs73486]

But, this would prevent all asset binary from getting replicated.

What, if  i need to replicate PDF assets fully to publisher?

[Jörg_Hoh]

I guess that then there is no real viable solution. Maybe the ReplicationPreProcessor can help here, haven't checked the details of it yet.

[venkateshs73486]

Hi Jorg,

Custom content builder with the following logic wouldn't help. Please suggest.

1. Pare the data

2. Identify if it is image/ non-image

3. Appropriately apply logic based on  asset type  . ( Replicate only metadata , if image and replicate full if not image)

[venkateshs73486]

To clairfy, i am asking whether Custom content builder with the following logic wouldn't help.

[Jörg_Hoh]

I don't think that a custom contentbuilder helps here, as you need to implement the binary transfer protocol ("durbo") to make it work with the standard replication endpoint /bin/receive. And the implementation of this protocol is not exposed via OSGI, but private to the replication bundle.

[venkateshs73486]

Thanks Jorg.

So, Agent user should have access to metadata node but not to renditions .

Is that the way it needs to be configured?

[venkateshs73486]

Thanks Borg.

Also, planning to check whether BL Replication serialization -type can replicate only metadata ,and not the binary, on publisher which does not external datastore.

[Jörg_Hoh]

Replication always makes sure that the complete node are available on the target system. If a shared datastore does exist, there is no need to transfer the binary anymore (and for this scenario there is the BL replication feature). But if the binary is not yet available in the datastore of the target system, it will be transferred. You cannot bypass this.

Jörg

[venkateshs73486]

Hi Borg,

One more thought.

Hope AEM allows to filter certain mime types during replication.

And i believe, we can use it on DMS7 mode.

Also, can you please share some document on how to retrict ACL access  to renditions folder but not to metadata folder of asset?

[Jörg_Hoh]

I am not aware to filter out mime types during replication.

Regarding ACLs: the term to look for is "wildcard acls" or "oak restrictions". A good documentation including examples is [1].

[1] Jackrabbit Oak – Restriction Management