Hi,
I’ve got a couple of questions about the slave scheduler.
[Edit: Where i say slave below, I probably mean ‘launcher’ since it appears to control the scheduling? Either way its all run as a service]
Our Environment
Our render farm environment includes our ‘artist systems’ which we have on a schedule between 11pm and 8am (non work hours). They are Win7 64Bit systems running the slave as a service in a Domain environment
Question 1
Is it possible (or feature request) to override the slave schedule per slave to disable it (handy for testing purposes or adding a system to the render farm when people are away or systems are sitting there empty for the day? IE turn it off without having to remove it from the machines group? Alternatively, could a colour or status be assigned to systems being managed on a schedule on the ‘slaves’ list in the monitor. Currently they just go grey when the slave shuts down, and its only if you know they are in a schedule group that you know why. Maybe they could be set to ‘scheduled’ and still be running, just not picking up jobs? It would also enable us to lookup the slave log while the systems are still online waiting until their scheduled start time.
Question 2
How often does a slave lookup its schedule settings from the repo? Does it only do it when the slave is online? From the looks of our setup (running as a service in win7) the slaves don’t boot up if the schedule group is disabled if the slave has shut down due to the schedule. Basically we have to wait until the next schedule start (11pm) or restart (actually, I dont think even a restart works…) the system so the service loads up and checks.
Another thing that’s happened for me is that I disabled the schedule while the group was in the ‘offline’ mode, yet there doesn’t appear to be any way to wake the slave back up again when its running as a service. The schedule had the ‘ensure slave is running during scheduled hours’ ticked, with a start time of 23:00 and stop time of 08:00 both ticked for all days. No other settings were ticked.
Fix for part of the problem - Looks like these systems were pulling an autoconfig to not start slaves on launcher start. Removed that and restarted and they appeared to start up. However, they still required a reboot.
Feature Request - Always running slave in schedule, disable job lookup outside of schedule
For slave scheduling, have an option to enabled the launcher and slave to be running and checking in with the repo/db but don’t allow them to process jobs until they are within the scheduled hours window. To me this would fix a lot the issues i’m experiencing, and would enable slaves to be pulled out of a schedule, enable viewing of a slave log outside of the scheduled hours window or have their pools and groups updated live without them sitting waiting for the next slave startup.
Feature Request - Scheduler: Option to start slave/or processing jobs if no user is logged; ignore users A,B,C,* etc
In our environment we can have high end systems sitting around for 1-6 hours while an artist (or anyone) isn’t logged into the system. This is during the ‘working day’ as well. We have all of our systems running the launcher as a service, so we can’t use (or trust) the artist to trigger the slave via the launcher.
For us, since we run as a service or might have other services running the ‘deadlineuser’ might be logged into the system and we would need a way to ignore it or any other users we want.
Hello,
Sorry for not getting you a reply here sooner. I will definitely pass along the results of what you found, and your feature requests. Just so you know, when a setting is changed on the repo it can take up to 10 minutes for the change to work it’s way to all the slaves, so it may not need a reboot, but instead needing just a bit of time.