Hi, which is the best practice to ingest data from external (internet-based) data sources, when only syslog or native forwarding are available? A set of load-balanced heavy forwarders in DMZ, that work as relay to internal indexers?
Direct channels from external networks to internal networks are not an option, due to security requirements.
As always, apart from the protocol, with any security-related zoning and such you have the question of the direction of the connection.
For some use cases it's ok to have a connection initiated by the UF "outside" connectiong to the HF "inside" (which is relatively safe even over internet if using mutual TLS - just make sure it's configured properly and working!) but for other cases it might be necessary to use some form of "pull" method instead of "push" - for this you might have to use some third party modules or even custom-written scripted/modular inputs to pull data from some external repository.
For syslog you could also use a setup with remote "collector" forwarding over an encrypted channel (like rsyslog with RELP) into your internal network where you'd forward the data into your splunk forwarders.
There are many possible setups depending on your limitations and restrictions.
unlucky I cannot move to a "pull" paradigm for these very specific data sources (network appliances). I use the "pull" paradigm with dedicated TAs (mostly of them were developed by me using the Add-on builder) for other cloud-based or external sources.
At the end I am investigating the use of a cloud-based log stream processor, which is able to forward data to Splunk using HEC.
Those both are usable solutions. Mostly it depends how you are collecting events on source side. If your are using syslog there then use it and if collection has done by UF/HF then just setup couple of IHFs/IUFs (intermediate xx forwarders) on your DMZ with TLS to get data into your site. And remember that use s2s is not (officially) supported via external LBs like F5/AWS NLB etc. Basically this means that you must use outputs.conf with FQDN names on source side.
thank you for pointing out that Splunk-to-Splunk forwarding is not officially supported via external network devices.
At the end I am investigating the use of a cloud-based log stream processor, which is able to forward data to Splunk using HEC. In this scenario, I would configure the reverse proxy to forward the HEC stream to existing load-balanced Heavy Forwarders, in internal network, where HEC input is active.