I would love to be able to have the dispatcher prefer certain machines when allocating jobs. For instance, always prefer renderfarm machines over workstations, etc.
Is this possible?
G
prefer blocks
prefer blocks
--
Rocky Linux 8.5, cgru 3.2.1
Rocky Linux 8.5, cgru 3.2.1
Re: prefer blocks
Hi.
You can tune renders priority in a farm setup.
You can tune renders priority in a farm setup.
Timur Hairulin
CGRU 3.3.1, Ubuntu 20.04, 22.04, MS Windows 10 (clients only).
CGRU 3.3.1, Ubuntu 20.04, 22.04, MS Windows 10 (clients only).
Re: prefer blocks
I'm not sure I follow, Timur,
Given the following hostmask setup:
workstation machines: lws[01-10]
renderblock b machines: rb[01-10]
renderblock c machines: rc[01-10]
video card block machines: rz[01-10]
I submit a job to hostmask: 'rz..|lws..'.
How would I signal the dispatcher to fill all the rz machines before dispatching to lws machines?
G
Given the following hostmask setup:
workstation machines: lws[01-10]
renderblock b machines: rb[01-10]
renderblock c machines: rc[01-10]
video card block machines: rz[01-10]
I submit a job to hostmask: 'rz..|lws..'.
How would I signal the dispatcher to fill all the rz machines before dispatching to lws machines?
G
--
Rocky Linux 8.5, cgru 3.2.1
Rocky Linux 8.5, cgru 3.2.1
Re: prefer blocks
Sorry.
Render priority is not in a farm setup, but can be adjusted from a GUI.
But all jobs will follow same render priority.
You can't setup one jobs to prefer one group of renders, and other jobs to prefer other.
For this, you should wait at least 2.3.x, that will not be in the near feature.
Render priority is not in a farm setup, but can be adjusted from a GUI.
But all jobs will follow same render priority.
You can't setup one jobs to prefer one group of renders, and other jobs to prefer other.
For this, you should wait at least 2.3.x, that will not be in the near feature.
Timur Hairulin
CGRU 3.3.1, Ubuntu 20.04, 22.04, MS Windows 10 (clients only).
CGRU 3.3.1, Ubuntu 20.04, 22.04, MS Windows 10 (clients only).
Re: prefer blocks
I'm glad to hear this is on your radar.
As most facilities (I imagine) we have a growing pool of (multi-gpu) GPU boxes some of which end up sitting idle if jobs get dispatched to the (much less desirable) workstations since they contain fewer GPUs.
I would still like to be able to utilize ALL the available resources, but only once the best resources have been consumed.
G
As most facilities (I imagine) we have a growing pool of (multi-gpu) GPU boxes some of which end up sitting idle if jobs get dispatched to the (much less desirable) workstations since they contain fewer GPUs.
I would still like to be able to utilize ALL the available resources, but only once the best resources have been consumed.
G
--
Rocky Linux 8.5, cgru 3.2.1
Rocky Linux 8.5, cgru 3.2.1
Re: prefer blocks
Code: Select all
How would I signal the dispatcher to fill all the rz machines before dispatching to lws machines?
Timur Hairulin
CGRU 3.3.1, Ubuntu 20.04, 22.04, MS Windows 10 (clients only).
CGRU 3.3.1, Ubuntu 20.04, 22.04, MS Windows 10 (clients only).
Re: prefer blocks
Ha! per machine priority is a lovely solution in this case... Thanks Timur, I hadn't thought of that.
As you can imagine, sometimes, when we are doing CPU renders (on the same machine groups) we'd want the big multi CPU machines to have higher priority.
Would you considers per-block machine priorities? Ug... feels... messy... there must be a better way! ;P
As you can imagine, sometimes, when we are doing CPU renders (on the same machine groups) we'd want the big multi CPU machines to have higher priority.
Would you considers per-block machine priorities? Ug... feels... messy... there must be a better way! ;P
--
Rocky Linux 8.5, cgru 3.2.1
Rocky Linux 8.5, cgru 3.2.1