Computational Resources Janitor

Inductiva API has at all times a janitor running in the background that will clean up any resources that are not being used. The focus of this janitor is to make sure that no computational resources are being wasted or left idle, for example, when a user forgets to terminate their machines.

Still, the janitor is mindful of the user’s work and will not terminate any computational resource right away. Instead, it will wait for a certain time before terminating the resource.

The janitor will also terminate any computational resource that is running for a certain time, for safeguards.

Moreover, to avoid cluttering the shared pool of resources, which is mostly meant for testing purposes, the janitor will terminate any task that is running also for a certain time.

The janitor follows the following rules before terminating the resources:

• Total time after launch of computational resource, independent of any simulation still running: 36h

• Total time of inactivity allowed, starting from the moment no simulations are on queue and reset if any arrive: 30 min

Please note that at the moment, the janitor is clumsy and doesn’t know how to handle the data of running simulations. So please be mindful of the time your simulations may take to run, and if this doesn’t suffice for your needs, please contact us.

The allowed machine group maximum validity (auto_terminate_ts) and the total time of inactivity allowed (max_idle_time) can be defined when initializing the machine group (this is valid for all machine group types).

import inductiva
import datetime

machine_group = inductiva.resources.MachineGroup(
    machine_type="c2-standard-16",
    data_disk_gb=20,
    max_idle_time=datetime.timedelta(minutes=1),
    auto_terminate_ts=datetime.datetime.now(datetime.timezone.utc) +
    datetime.timedelta(hours=10),
)

When omitted, these parameters will have the values mentioned above, a timestamp that is 36 hours in the future for auto_terminate_ts and 30 minutes for max_idle_time.