Hi,

I was able to install the HQueue client and server and get them up and running. I added the hq_render node after a mantra render node, then clicked ‘Submit Job’. The job seems to submit fine as I see a dialog that says "Your job has been submitted (Job 29). <Open HQueue> <Ok>.

Clicking <OK> closes the dialog. Clicking <Open Hqueue> and get a no apps installed to run localhost.

So, I don't know how to get the jobs to start rendering, nor do I know how to see the jobs in the queue. I have looked at the documentation for HQueue, but did not see anything that seemed to solve this.

Thanks,

Adnan

Hi Adnan,

It just sounds like the Open HQueue button is foo-barred for some reason.

But you should still be able to manually open a web browser and point it to your HQueue server to view the jobs (i.e. http://myServerName:5000). [myservername] Then you can see if your jobs are succeeding or failing and why.

Cheers,
Rob

Thanks for the reply. I was able to open the queue monitor via the web browser thanks to your help.

Jobs were not rendering because Hqueue can't render takes.

It seems, despite the FAQ, there are things that are difficult to figure out regarding Hqueue, so I made a little page with some info as I explore further.

http://www.madguru.com/learn/working-with-houdini-hqueue-windows [madguru.com]

Hi Madguru,
thanks for the webpage…I actually had the problem with the IFD files and Indie that you mention so it was of great help. I was wondering if you have problems of render speeds on Hqueue as I do. It feel the Hqueue renders like x5 times slower than rendering directly from Houdini on my set up. Everything else works great. just very slow the render and the hython preparation of the scene (maybe this is an Indie issue?)

I'm glad I could help with the IFD files problem. I haven't had a problem with render speeds on HQueue, and I am also using Indie. It may not be using all the processers. Double check if this is the case. In your render node ‘hq_render’ from which you clicked submit, go to the ‘Advanced’ tab and check if the ‘Set Number of CPUs Per Job’ option. Make sure it is using as many CPUs as you expect.

Hope that helps you get those renders faster.

Nicolas Heluani
I was wondering if you have problems of render speeds on Hqueue as I do. It feel the Hqueue renders like x5 times slower than rendering directly from Houdini on my set up. Everything else works great. just very slow the render and the hython preparation of the scene (maybe this is an Indie issue?)

Hi Nicolas,

Are the jobs slow to startup? (i.e. slow for the jobs to go from the “waiting for machine” status to “running” status). Or are they slow when rendering?

If startup is slow, then it could be a DNS resolution or general network issue. When a new job is submitted to the farm, the HQueue server notifies the client machine that a new job is ready for processing. If the server is unable to connect to the client for whatever reason, then the job sits around “waiting for machine” until the client contacts the server during its regular ping interval (typically between 3 to 5 minutes).

The server may not be able to contact the client for any number of reasons but the most common issues are a firewall running on the client or the client's hostname not registered with DNS.

Now if rendering is slow, then check the “Set Number of CPUs Per Job” option as suggested by madguru and set it to a higher number. Or you can also uncheck the option and Mantra will utilize all CPUs.

Cheers,
Rob

rvinluan
Hi Nicolas,

Are the jobs slow to startup? (i.e. slow for the jobs to go from the “waiting for machine” status to “running” status). Or are they slow when rendering?…

Hi Rob, and Madguru.
First a quick note: When I talk about client let me be clear that this is the same machine for Server and client. Later I plan to grow it, but for now I started with one.

The jobs are slow at the preparing the scene task, around 10-16 minutes,and the hython never really uses more than 1% of my CPU. And once it is rendering is about x4 to x5 times slower than just using render to disk option BUT…

Your posts helped me out in guiding me where to dig around for the problem. At first I thought it was the CPU issue, but the Set Number of CPUs Per Job option was properly unchecked. And again once the Mantra process starts it run fast and complete the frame without issues, but it is constantly turning off an on. So the problem it seems is the division of each frame as a different job, even though I only have a single client/server (my current machine) each frame is a single individual task and each time the frame is rendered (which happens fast) there is a lot of wasted time opening the new frame. The same for the preparing the scene task. The task itself is very slow according to how many frames have to prepare in individual jobs.

So all the speed problems goes away once I either increase the Frames per job or check Batch all Frames in One Job.
On a side note this was a very fast render (12 seconds per frame), maybe this dead time becomes negligible in more heavy renders, but I do feel this solution is kind of a work around rather than a solution in itself.

PS: sorry for the long post, I try to be as precise as possible.

Hey, not sure what OS you're using, but I found that button doesn't work with the Cinnamon desktop. I logged a bug, and it was because Cinnamon/Muffin (Linux Mint) handled that event different from other Linux desktops.

I've also found it more reliable to add .local. The server is on a different machine of the network, but wrench.local:5000 is the url for the monitor, and the ROPs also point to wrench.local:5000.

goldleaf
Hey, not sure what OS you're using, but I found that button doesn't work with the Cinnamon desktop. I logged a bug, and it was because Cinnamon/Muffin (Linux Mint) handled that event different from other Linux desktops.

I've also found it more reliable to add .local. The server is on a different machine of the network, but wrench.local:5000 is the url for the monitor, and the ROPs also point to wrench.local:5000.

I´m using windows 10 (I should have mention this at some point, oops) but like I said everything in my setup is working appropiately is just that the client takes too long opening each individual task inside a job.

And again once the Mantra process starts it runs fast and completes the frame without issues, but it is constantly turning on an off. So it seems the problem is the division of each frame as a different task, even though I only have a single client/server (my current machine) each frame is a single individual task and each time the frame is rendered (which happens fast) there is a lot of wasted time opening the new frame. The same for the preparing the scene task. The task itself is very slow according to how many frames Hqueue have to prepare in individual jobs.