Render Farm Open Issues
- How can alfred use render servers that are not in the OUGL lab?
- 24-Apr. DWJ. Tim has successfully dispatched jobs that the maitre_d (on darjeeling) assigned to rurou and nerra
for remote rendering!
- 21-Apr. DWJ. Rurou and Nerra are installed and running in my office. We will use them this week to explore getting remote render machines enabled.
- 18-Apr. DWJ. My office machine is too slow for our needs. However, both of the old machines from the lab (rurou and nerra) have been moved from the lab to my office and will be hooked up on Monday 21-Apr as prototype render slaves.
- 17-Apr. DWJ. alfserver will be installed today or tomorrow on Doug's machine in Sieg 226D for Tim to experiment with.
- Some machines show as available (active alfserver) but when you try to render they show as not available.
- 5-May. DWJ. This is still an issue. We have increased logging levels for maitred and will do so for the alfservers.
- 24-Apr. Ben Nordstrom- "last night we attempted to render 360 images for a short sequence.
the render seemed to start fine, however, all of the fastest machines in
the labs were un to render. they gave errors such as "metric not up to
date" and "alfred not available" or something similar."
- 24-Apr. Investigation delayed while basic connectivity issues were worked.
- 17-Apr. CL/DWJ will investigate.
- 17-Apr. CL - "This morning I tried a render and found that about half the lab computers had the same "no alfserver" message."
- What is the policy for rebuilding?
- 22-Apr. DWJ. The rebuild time has been changed to Friday night (Sat 3AM). We will encourage everyone to log off when
they leave on Friday, and the machines will rebuild overnight and be ready Saturday morning.
- 17-Apr. DWJ, per SS: Here's the rebuild policy, at least what's in effect now: machines are rebuilt once a week,
around 3am Sunday morning. This time was selected by Barbara last fall, in an effort to balance machine stability and minimum
user inconvenience. Overnight renders shouldn't be started on Saturday night, but are A-OK any other night of the week.
- What is the plan for the new machines?
- 5-May. DWJ. All the machines in Sieg 228 have been added to the farm. We are working to include the 10 Linux machines that
belong to the Steve Seitz and the Grail group as added render servers.
- 18-Apr. DWJ. The five new machines (alpha, chris, gumpa, elder, dopey) are installed in the lab.
The old machines ugeso, stode, and zibbu will stay in the lab as headless renderers.
The old machines rurou and nerra are in Sieg 226D as prototype remote renderers.
- 17-Apr. DWJ. There are five new machines coming soon. They will replace five existing machines in the lab.
Three of the old machines will stay in the lab as headless render slaves, and two of the old machines will
go to Sieg to be test beds for remote render slaves.
- We need html access to maitre_d status and convenient links to it.
- 1-May. Per SS: "It's still being worked on."
- 22-Apr. SS. I did a little today, and realized that I needed assistance from Pixar, so
submitted a support request to them. The 'enabling the wrangler mode'
stuff in the documentation is UNIX-flavored, and doesn't apply to Windows
installations.
- 17-Apr. SS is working this.
- We need some tutorials with screen shots showing how to use alfred in various modes.
- 23-Apr. Chad's tutorial is linked on the Alfred page. Comments or problems - send to cse490ap-render@cs.
- 21-Apr. Chad will have first draft on Wednesday, 23-Apr.
- 18-Apr. Chad is working on this. The proposed outline is:
- A simple rendering job that uses all machines.
- A multi-job example started from one machine.
- A multi-job example started from several machines at the same time.
- When we get the machines in 228 linked up to run as part of the render farm, we will be using
all the licenses that we currently have available for running alfred (40). How many other machines
in which labs do we plan to use?
- 24-Apr. DWJ. Plan in work.
