This page focuses on security considerations of running agents.
The security considerations for the overall Replicante ecosystem design are documented in the security section of the admin manual.
By default communication with replicante core happens over HTTP, with core initiating connections.
This is not a secure setup (unless you trust the network, but you don’t right?). At the very least this exposes the system to replay attacks: a malicious user can record a legitimate request and re-send it to the agent at will.
Offical agents support HTTPS-only servers, with mutual certificte verifiction required for actions to be enabled.
The actions system can only be enabled if mutual HTTPS verification is enforced.
This is by design as actions can easily compromise datastores if used casually or maliciously.
The agent may require some privileges on a server to perform actions such as restart datastores or update TLS certificates.
The recommended approach is to run the agent under a user with limited permissions and grant extra permissions as and when required.
Specific agent’s documentation will provide extra information and details regarding the permissions needed by the agent itself.