Reduce the number of restarts when deploying to clients using deployment server

[imacdonald2]

At the back of my mind I remember reading something about reducing the need to do restarts when deploying apps. Can someone point me to the documentation about that? Does the deployment client automatically know to do a restart? For example do you need to do a restart if you add a new index?

When do you need to do

restartSplunkd=true

Most of my apps are simple input apps that have indexes, inputs and props and thats about it.

[dwaddle · Accepted solution]

I set restartSplunkd=true on every app I deploy to forwarders -- mainly due to the fact that 99% of the configuration elements that forwarders use require a restart to take effect. When deploying to indexers / search heads it's a little less cut + dry. Adding indexes and inputs (as of Splunk 4.2) definitely fall into the "must restart" category.

[dwaddle]

In my deployment, I don't deploy to search heads (right now). But, it does make sense that it would blow away <app>/local when deploying a new version. If it didn't, then you couldn't be sure the new updates took effect properly. The whole idea of deployment server is uniformity, and letting each search head override in local is probably not the best choice.

[imacdonald2]

I am curious about you saying you deploy to search heads, I found that the deployment server wipes users settings in app/local/ when it deploys a new version.