Keep history for a chosen duration
Set your node in inventories/example.yml, then run:
mmx run configure
mmx status
mmx results --json
run_history_retention: 24h is the default. Choose another positive
duration such as 7d, or forever. The setting governs history on the
controller machine.
Nodes keep each finished job's result for 72 hours, so a controller that
lost contact can still recover the result with mmx status or mmx resume.
After 72 hours, ordinary MMX traffic removes the node copy. Queued and
running jobs stay.
mmx results reads the controller's copy. Recorded resume uses the
controller's copy and gives unfinished work a new job ID. mmx results and
recorded resume continue to work after node expiry while the controller copy
is retained. Controller cleanup keeps the latest normal run for this
project/playbook/inventory selection, even when older than retention. Node
cleanup leaves service hashes and installed runner binaries alone.
The acceptance suite exercises node expiry, disconnection, and explicit resume on disposable nodes.
This page is examples/history/README.md in the MMX repository.