Execution modes and processes#
There are several different modes in which you can configure n8n to operate when executing workflows. These settings become more important as you increase the number, and complexity, of your automation workflows. Ensuring you set n8n to run in the most beneficial mode is also crucial when you begin to scale n8n to handle very large workloads.
The mode in which you n8n instance will operate is set via the following two environment variables:
||Determines whether each execution runs in its own process, or all run in the main process.|
||Determines whether executions are processed by a single instance or are distributed across worker instances.|
There are two options available when selecting how execution processes are handled in n8n:
This is the default setting in n8n. In
own mode, each execution runs in its own, newly instantiated process (i.e. 1 execution = 1 running process). Running in this mode has the following benefits and disadvantages to consider:
|Stability: one crashed execution does not impact others.||Latency, approximately 1sec as each process spins up.|
|Efficiency: all CPUs are utilized.||Resources (CPU/RAM) needed for the additional processes.|
This mode is recommended for running CPU intensive tasks to ensure processes do not block each other.
own also reloads nodes on each execution, it is also the best mode for developing custom nodes.
main mode, all executions will run in the main n8n process. This mode has the following benefits and disadvantages to consider:
|Minimal latency||Cannot take advantage of multiple CPUs. Single process can result in a bottleneck.|
|Resource efficiency: only one CPU required.||Reduced stability, one crashed executions causes all others to fail.|
There are two options available when setting the executions mode:
This is the default setting for n8n. When running n8n in the
regular mode, everything is processed on a single instance. Executions are handled according to the
EXECUTIONS_PROCESS setting and available resources.
Queue mode is designed for handling very high workloads. In this mode you run multiple instances of n8n: one ‘main' instance coordinates, and other ‘worker' instances actually process the executions. You can also add dedicated instances to handle incoming webhooks.
Queue mode requires more setup than
regular mode, but provides great scalability and stability with minimal latency.
To learn more see the Scaling n8n documentation.