4

Let's say I have a 2013 SP1 system with a fairly large BP (50 sites) with high volume that suffers from poor performance in publishing & deployment (unscaled) with all content going to a Tridion Broker DB (DD4T TBBs)

I am planning on suggesting the following

  • Increase the number of dedicated content publishers, separate machines
  • Increase the number of deployers, separate machines, load balanced

Now, as I understand perhaps the best way do this is to split the publications so each one is assigned uniquely to a specific publisher for optimum publishing. Ideally I'd like to do that with a single target of "Production" to avoid separately named targets, but I can probably live with that by limiting the publications per target so there's no cross pollination.

However I'm not sure how to best configure the Deployer architecture - will there be consistency issues in the DB and with CCS if there are load balanced deployers writing to a single DB?

3

Following Julian's post you would create multiple publication targets each with it's own set of publications. Then associate them with a single target type so the content editor only sees one logical target. The deployer configurations could all be the same.

0

For deployers you will have to go for failover configuration that is you have two deployers and only one deployer is active at a time.

The problem with making deployers load balanced is: After package is sent from publisher to deployer, publisher keeps polling deployer to know the deployment status and in case of load balanced package may be sent to one deployer which is doing deployment and polling is done to other deployer so it will never receive the correct status and eventually fail.

You can refer to following question to learn about how to setup active/passive (failover) deployers: Active/Passive deployer setup

  • Interesting point, can you possibly deal with this by sticky sessions on the load balancer, so the route is bound on first connect? – mpaton Oct 22 '15 at 7:42
  • 1
    There is one work around that we are trying to evaluate: is to have shared location for both deployers so in case of failover the shared location still has the package and processing can continue. However, need to test all different scenarios. – Hiren Kaku Oct 22 '15 at 14:50
0

You can also try increasing the WindowSize and Workers of the deployer in cd_deployer_conf.xml

For example, the default is:

Location Path="c:\tridion\incoming" WindowSize="20" Workers="10" Cleanup="true" Interval="2s" RetriesInterval="3s"

Modified:

Location Path="c:\tridion\incoming" WindowSize="50" Workers="20" Cleanup="true" Interval="2s" RetriesInterval="3s"

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.