Render Pools

[zaren]

It can be like in QubeVFX:

http://docs.pipelinefx.com/display/QUBE ... b+Priority

[lithorus]

If create some Pools, lets create a useful Pools!

We can also make use it to split farm between departments and projects.
If we add a some job "Pools" too (along with render pools).
If we add priorities to such pools, we can adjust projects priorities (give some project more hosts).

A project job attribute would definately be good to have and later project attribute for render hosts.

[seven11]

Hi.
The same i can do with a mask parameter, which already exists:

Code: Select all

"mask":"stationa|stationb|stationc"

or:

Code: Select all

"mask":"station[a|b|c]"

Hi Timur,
I want to try this out. What file should this get placed into?
Thanks,
Scott

[timurhai]

Hi.
In a farm setup to limit services:
https://github.com/CGRU/cgru/blob/maste ... ample.json
In a job/user hosts mask.

[lithorus]

After seeing that the render pools have been implemented I wanted to give it a spin.

I would REALLY miss the farm.json setup. Not having the ability to dynamically assign capacity, services through rules is a BIG loss IMHO. Wouldn't it be better to have rules to assign pools? I don't see the benefit of having a GUI based ressource assignments.

Also is there any way to save the state of a render farm? I can clearly see a renderfarm setup messed up by a random user clicking around in GOD mode.

And this is such a huge change that it should go in a 3.0 release.

[timurhai]

Hi.

For now pools implementation is not finished.
A pool has a "pattern" parameter.
New hosts will be parented to pool matching pattern mask.
So i do not see anything that we losing from a farm.json.
But now you can move some host in any pool manually, it is useful to setup specific servers.

Yes this is a GUI based resource assignment. But any GUI communicates server via json protocol, so you can integrate pools parameters changing in your pipeline.

You can't save pools state like you can just copy farm.json. But any nodes state server stores in its json store.

Also for now there are no service limits. But it is just not implemented. It will be soon.

[lithorus]

Come to think of it, one could make a farm.json parser which applies the rules on the farm. If this can be run at startup it would still be compatible with the old setup.

I do like the idea that one can add and remove services on-the-fly (which is actually a problem I'm currently facing). Although I would like to be able to lock it down to certain hosts.

I can work on implementing the service assigning in afcmd.py.

Hmm.. is capacity also gone? Or still needs implementation?

Again, this should be a 3.0 release IMHO.

[timurhai]

...
I do like the idea that one can add and remove services on-the-fly (which is actually a problem I'm currently facing). Although I would like to be able to lock it down to certain hosts.
...

- Now we can add/remove service for a permanent purposes and disable/enable for a temporal.
Capacity is implemented. By default it is -1 (no limit), but limited a number of tasks per host (1 per host).

[timurhai]

And yes it can be 3.0, i thought about it too.

[lithorus]

Capacity is implemented. By default it is -1 (no limit), but limited a number of tasks per host (1 per host).

Where can I set it? I can't seem to find it in the GUI.