Hi,
We are attempting to power down our renderfarm and servers and was wondering what is the best way to stop our renders and start them back up again without disturbing other farm jobs that are stopped. Is there a way bring them back in the state they once were but only for jobs that are running/pending/queued?
Thanks
Usually people would use the “Stop Slave after Current Task” remote control option.
That’ll require either valid DNS for your machines or “Use Slave’s IP Address for Remote Control”
The nice thing here is that the jobs won’t change in any way, and that option will allow the farm to shut down gracefully instead of just pulling the plug on all the renders.
thanks for that but what about the renderfarm server / repo server itself? how do we shutdown the server after the slaves are dead and then bring it up again?
thanks
Deadline has no central server application officially. The closest things are the MongoDB Database and the Remote Connection Server (which isn’t used by everyone).
Deadline was engineered to handle when those unexpectedly disappear (power failure, machine reboot) so you can just stop and start them. Info on stopping and starting the Database is here.
If your talking about restarting the machine, just reboot it. The Slaves will wait until it comes back, including those that are rendering. If not, I’d classify that as a bug.
So if we have a deadline pulse running on a different server, do we shutdown that before the repo server?
So is the sequence for shutting down the farm, turn off the slave machines first.
Then turn off deadline pulse server
Then turn off the repo server?
is that correct?
thanks
Yup. That’s the best.
Honestly, I’ve restarted each of those components at various times and not had issues and that is how it should work for the core components. AWS Portal components can be a little touchy, but even those are designed and tested to go missing and come back at unexpected times.
Since we have had to do this a pile of times, due to being in a shared building.
Disable all the slaves, stopping the slave.
We stop the pulse/ webservice nodes.
Once the slaves/pulse are stopped, set the machines to shutdown, this is done out side of deadline.
Gracefully stop the mongodb service on the server and shut it down.
That usually does the trick and causes the least things to be out of sync.
We bring up/ online in the reverse order.
1 Like