Qmaster Cluster just doesn't work?

Yes, it does appear that your Cluster is running. And yes you don't need the Qadmin app for now. It is odd that Compressor can't see the cluster since Qadmin can.

That's where we differ. My Quickcluster shows up in the list and I can't even delete it if I try.

I just stopped sharing my cluster in the middle of a render and reshared it. My cluster is now in the same shape as yours at the beginning of this thread. Compressor can't see the cluster and neither can Qadmin. I have to head home, but will pick this up tomorrow.

Any luck? I rebooted the Quickcluster Mac this morning. Reset the Cluster (option-click), and re-shared. Booted the service Macs and I'm back on. So I'm thinking that something got reset by rebooting. The question is what...

Well, I just confirmed it for myself. FCP must be present on each Mac for distributed Compression with Compressor. I got hoodwinked. Our IT guy installed the full FCP studio. When we told him that we didn't have enough licenses he made his move. Not knowing how to uninstall and too scared to delete them (PC guy), he moved all the pro apps into another folder inside Application folder. He then installed Qmaster on all the Macs. You can guess the rest...someone is looking for a lesson on uninstalling Pro apps.

If I drag even one FCP app in the cluster's group to the trash and try to render, the cluster fails. Put the app icon back and the cluster renders. Therefore, there must be some encoding or network resources which are in FCP.

Sorry, for all the extra fuss that you when through. Hopefully, Apple will move these resources out of the FCP install and move them into the Qmaster install.

Cheers,

AV

You can get around the FCP on every node thing by using reference movies. Just make sure all the assets are on your shared volume, and make sure the reference movie is as well. Then submit from Compressor, not FCP.

Otherwise yeah, you need FCP on all the machines if you submit from FCP. Kinda *****.

Hi,

I don't understand you last post. Did you ultimately determine that the same license of FCP can be used on multiple machines for cluster rendering (like the render engine in After Effects) or do you need unique licenses for each machine? That would be very expensive.

I don't know if you're having the same problem I had, but I just got my cluster to work finally, on four machines and it is fantastic. If youare not seeing the computer after having stopped sharing, or changed in the middle of a project, then try this.

Stop sharing on whatever machine is not being recognized by the cluster, then hold the "option" key and click "reset services." It only shows up in the system preferences when the option key is held down. (Start sharing switches to reset services.) Then stasrt sharing again and look at the cluster in QAdmin.

Also, quick clusters in my opinion are not as good as a cluster set up in Admin with a controller and nodes strictly set up for services only.

If your thing doesn't work after resetting services let me know and I will tellyou my exact setup. It took a week to find it all out but after about twenty questions on three different forums and having read each manual three times, Compressor, Qmaster, and Qadministrator (all have al,most the same manual by the way) I figured out how to do it. I will talk you through each step if you need.

Red Truck

I think I've got qmaster set up correctly. Compressor sees all the machines, qadministrator shows the job going to each of them and I do get a finished qt movie. Problem is it looks like most of the machines spend the majority of the render idle. There seems to be no rhyme or reason as to why or when they go from rendering to idle. Could this have to do with the cluster controller also being part of the render? Could it be that the CPU is sufficiently taxed that when it comes time to dish out the next round of data to waiting clients it is too tied up rendering to respond? I've got a dual 2.0ghz G5 as the main machine with the 1000baseT accessible Xserve with all the source material, its also the FCP machine. There is a 733mhz G4, a 2.0ghz core duo iMac, a 2.0ghz core duo Mac Book pro and a 1.5ghz core solo mac mini. The intel machines are usually as fast or faster than the dual G5, but they are sitting idle for most of the render. Any ideas?

Are they idle because the segments are done? What happens is the job is split up into segments (2 x number of nodes), and each node does a segment, and when it finishes it moves to another segment. So if you do not have another batch waiting in the queue, nodes will sit if there are no free segments. If there is another batch, the free nodes move to the next batch. Looking at your machines in your cluster, I'm guessing that you G4 733 is processing, and the rest are idle. this would be because the other machines are at least 2-3x as fast as that machine, and it will take a long time to do its segment. You should remove the G4 from the cluster.

You can view what machine does what segment in Qadministrator. Also, if you expand the batch tree in Batch Monitor, you can see how many segments, what is finished, and what is processing, and what hasnt been done. I would think that you probably have 1 segment thats processing (on the G4), and a bunch that finished.