How to run both Control Plane and Data Plane on the same machine with DNF/RPM installation in decoupled mode? #12874
Replies: 1 comment 2 replies
-
|
Hi @AjithNirmal! I'm Dosu and I’m helping the apisix team. Yes, you can run both the Control Plane (CP) and Data Plane (DP) on the same machine using the RPM/DNF installation—this is supported as long as you manage configuration, ports, and systemd units carefully. Here’s how you can do it:
Caveats:
For more details and config examples, see this GitHub issue. To reply, just mention @dosu. How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other
|
Beta Was this translation helpful? Give feedback.
-
|
Is this approach recommended or battle-tested in the community/production environments? Specifically: Sharing the same base directories (like /etc/apisix for configs, /usr/local/apisix or default paths for logs/PID if not overridden)? I understand the official docs emphasize separating CP and DP instances (often on different nodes/containers), is this dual-instance-on-one-machine pattern commonly used and stable? Any known issues with resource contention, port conflicts, or config syncing in such a co-located setup? |
Beta Was this translation helpful? Give feedback.
-
|
This dual-instance (CP + DP) pattern on a single host is technically supported, but it's not a common or battle-tested practice in production—there are no official recommendations or community case studies for it. The official docs and most real-world deployments emphasize separating CP and DP on different nodes or containers for scalability, security, and operational clarity docs. If you do run both on one machine, you must use separate config files, assign unique ports for admin/proxy, and—critically—set different log and PID file paths for each instance. The default RPM/DNF install and APISIX CLI hardcode There are no built-in limitations preventing this setup, but you are responsible for avoiding port, log, and PID conflicts. Resource contention is possible if both planes are busy, and you should monitor system resources closely. There are no known issues with config syncing as long as both instances point to the same etcd cluster and have correct roles set. In summary: this approach is fine for testing or small environments, but not widely used or recommended for production. If you proceed, be meticulous with config separation and resource allocation. To reply, just mention @dosu. How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other
|
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Hi team,
I'm setting up Apache APISIX in decoupled deployment mode using the official DNF/RPM package installation on a RHEL-based system.
The standard DNF installation creates a single APISIX instance with:
One default config.yaml in /etc/apisix/
One systemd service (apisix.service)
In decoupled mode, I understand that the control plane (CP) and data plane (DP) need to be separate instances, each with their own configuration (one set to role: control_plane and the other to role: data_plane), sharing the same etcd cluster.
My goal is to run both CP and DP on the same physical machine (for testing or a small setup), while sticking to the RPM/DNF installation method.
Questions:
Is this supported/recommended?
How can I achieve this? For example:
Do I need to duplicate the config files (e.g., config-cp.yaml and config-dp.yaml)?
Create separate systemd services (e.g., apisix-cp.service and apisix-dp.service) pointing to different configs?
Handle port conflicts (e.g., different proxy/admin listen ports) and separate log/PID directories?
The official docs focus more on Docker/Kubernetes/Helm deployments for decoupled mode, but I'd prefer to use the native RPM package if possible.
Any guidance, examples, or pointers to similar setups would be greatly appreciated!
Thanks,
Ajith Nirmal
Beta Was this translation helpful? Give feedback.
All reactions