Deactivate assets

[kevingtan]

Hi All,

We are running AEM 6.0. We only set up Author-to-Pub and Pub-to-Web agents, but there is no invalidate flush agent. The current setups are not good for assets such as images and PDF's. The problem is that whenever an asset is deactivated or modified, we cannot see the asset having the correspondent action, either deleted or updated. I figured that we are checking the "On Receive" option on the "Triggers" tab, see attached.

I like to have that option checked, but concern about the subsequent impact on the dispatcher server. Any idea of the dis/advantages checking that? Or do I have another way to the deactivation more responsive without a significant impact on the dispatcher compared to the current settings?

Thanks!

[Gaurav-Behl]

What is "Web" in this case? Is it AEM publish server or Apache server?  Do you have the following setup:

Author > Pub1 > Pub2  > Web server (Dispatcher)

Is "On Receive" turned on on Pub1? Do you mean to day that you don't have any dispatcher flush agent/invalidation agent configured on either servers?

[kevingtan]

"Web" is the dispatcher. We do not have the setup as you laid out. We have the following setups:

Author:

    Agents on author:

        Author-to-publisher

    Agents on publisher:

        No active agents

Publisher:

    Agents on author:

        No active agents

    Agents on publisher:

        Publisher-to-dispatcher

The attached image is the "Publisher-to-dispatcher" on the publisher. Yes, we don't have any flush agent on either server.

[Gaurav-Behl]

I suspect that "Publisher-to-dispatcher" agent is your dispatcher flush agent.

What is the "Serialization Type" of "Settings" tab of "Publisher-to-dispatcher" agent?

Do you see HTTP headers configured in Extended tab?

In any case, the purpose of turning-on "On Receive" is that this agent (publisher-to-dispatcher) would trigger automatically as soon as it receives any replication event from author server

Similar explanation goes for "On-modify". When a content author would modify anything on this publish server (which doesn't make sense because of lack of authoring capabilities), this agent would trigger.

"On-Modify" makes sense on author and not on publish servers unless there is a specific use case. On author, it would trigger replication event as soon as the modified content is saved.

I'm not aware of the purpose of keeping a "replication agent" on publish server instead of "flush/invalidation" agent. To me, it doesn't make sense. Probably, you have a TTL based cache flush setup rather than an event based cache flush setup.

[kevingtan]

Thanks for your explanation. The serialization type of the agent is "Dispatcher Flush".

These are the settings for "Settings" and "Extended" tabs.

So your suggestion is that we should set this up in Author instead of Publisher?

Thanks again!

[Gaurav-Behl]

Correct, this is flush/invalidation agent.

Adobe recommends to setup flush agent on author rather than publish servers unless there is a specific use case related to time lag, race condition or some business/technical use case. It would still function in the same way.

What is the exact issue that you observe with this setup? Is it related to stale cache or that imgs/pdfs do not flush but html files flush out or the time taken to flush cache or performance related issues or something else?  Could you please elaborate on  "The current setups are not good for assets such as images and PDF's. The problem is that whenever an asset is deactivated or modified, we cannot see the asset having the correspondent action, either deleted or updated."

the firewall rule must be because of DMZ..

[kevingtan]

Well, I think we figured we should implement the way as you are suggesting. Thanks a lot for your help. In regards to your question,  it means that without the "On Received" checked, the cached version of the asset is always there regardless of what you do on the Author, even if you deactivate the asset and upload/activate a modified version of the asset. But this won't happen to the HTML pages.

[Gaurav-Behl]

"On Received" is typically enabled when the flush agent is setup on publish (not on author) to trigger that agent in context. When replication is triggered on author (by any means), it triggers a flush request as well. If the flush agent is not setup on author then this request doesn't perform any action due to the absence of its handler.

"On Received" should be enabled on publish server because the publish server doesn't know when to trigger flush request for that resource/asset apart from listening to incoming replication requests. On author, it wouldn't help because the implicit flush request would serve the purpose.

[kevingtan]

Also, we set this up probably we have a firewall rule that only the Publisher can communicate with the Dispatcher.