I would recommend you upload form the search head, but you need to ensure you confirm that the search head is configured with an ouputs.conf and is forwarding to the indexers. It is best practice, that all your splunk instances other than indexers* have a outputs.conf and forward their logs or any data uploaded to the indexers.
You can read the following article and think of your search head as a "Heavy Forwarder":
http://docs.splunk.com/Documentation/Splunk/6.6.2/Forwarding/Deployaheavyforwarder
As long as you have an outputs.conf on the search head, uploading form there will be fine.
Now, when you upload the file, it will be sent to ONE indexer, indexed, and replicated based on your Replication Factor/Search Factor in the cluster.
When you say "all indexers", do really mean every single indexer? Are we talking about a file that needs to be indexed? or a file that needs to be provided to each indexer...Because if it's the later, you simply need to send the file to the indexers in an app from the Cluster Master by pushing a cluster bundle.
http://docs.splunk.com/Documentation/Splunk/6.6.2/Indexer/Updatepeerconfigurations
- there are certain scenarios where indexers might have outputs to 3rd party systems but thats beyond the scope of this answer
- MattyMo
