I have been using Hqueue heavily the past two months. As the sole Houdini user in our facility I do my best to maximize our resources. I have two dedicated render clients and my laptop which I develop on. For what it is worth these are the issues I have struggled with. Would be grateful for helpful insights anyone could offer.
#1. Not an HQ issue but I cant use shared folders on our corporate net. I have to resort to manually copying folders to each machine. Is it possible that I could share Teams/Sharepoint or other folders for the client machines?
#2. With some scenes while generating the usd, the machine accumulates memory and tops out and crashes.
- can this job be restarted at the usd generation level? if I reschedule it seem to restart from the beginning clobbering all of the other frames and likely crashing at the same time location.
#3. it would be REALLY helpful to have more relevant info compressed into the main ui. Most if not all of the relevant job and client info should be available from the main screen. ie being able to see and select which machines are working on a job. Which machine is holding the FX or Engine licenses... Switching between the client and job ui are cumbersome and slow as well as not conveying relevant job info.
#4 Licensing. I have an FX license and only recently went to the web license mode which is a HUGE HUGE help. With my current problem from issue #1 I need for each machine to generate its own usd file for Karma rendering but often a machine seems to keep its FX license even though it is only doing the Karma job and I have to manually release it. Ugh. Again the license details would be great to have in the main Ui of Hqueue.
As with most of my Houdini questions I am often humbled by how simple the workflow or solution can be.
Apologies for the long thread. Just looking for workflow suggestions and insights.
JeffMc
Found 123 posts.
Search results Show results as topic list.
Technical Discussion » hqueue wishlist
- jeffMc
- 125 posts
- Offline
Technical Discussion » Rescheduling failed renders in Hqueue
- jeffMc
- 125 posts
- Offline
I have Hqueue up and running. If a job fails part way through, is there a way to restart it and have it not clobber the existing valid files in the frame sequence? I am referencing both the .usd generation and Karma rendering. thank you Jeff
Edited by jeffMc - Feb. 27, 2024 20:45:19
Technical Discussion » render pipeline options without a 'farm'
- jeffMc
- 125 posts
- Offline
Just a follow up on this. I realize now that with the login licensing that I can submit a usd job to one of the render clients and if I close my workstation license quickly the render client can grab it to generate the usd files and then go back to a Karma license when finished. This really helps my workflow since I only have a single FX license to work with. Coming from RoyalRender HQueue feels a little primitive but it gets the job done.
Technical Discussion » render pipeline options without a 'farm'
- jeffMc
- 125 posts
- Offline
Tanto
Can't solve your whole problem, but just to let you know that you could generate your usd files on your work machine and render it on your other ones with Husk, using only Karma licenses. No need for Engine licenses if you do it that way.
Thanks so much for this. I have managed to get the machines working by manually copying the generated usd across the network to a commonly named folder. I can then use HQueue to manage the Karma rendering on them. woohoo
Jeff
Technical Discussion » render pipeline options without a 'farm'
- jeffMc
- 125 posts
- Offline
I work in a corporate IT environment and I have 3 or so very nice machines that I could use as render slaves. However they are trapped behind a strict corporate firewall. Being unable to get my IT to open ports or shared folders which I believe any render farm app would need I am looking for options. My efforts to set up Hqueue have been futile and I have some impending deadlines.
My scenes use Karma and I have a single FX license which I believe provides up to 3 Karma render nodes.
I don't mind doing command line renders on each machine if that is what it takes.
What is my best and most cost effective options to employ these machines for rendering my scenes?
Would I need an Engine License for each machine that I need to do a command line render on? $749 per machine?
I am familiar with generating ifd's for Mantra renders and it seems that generating the .usd for Karma is a similar workflow and I am good with that.
My last options seems to be to ditch these machines and turn to a cloud rendering service. Open for suggestions on good Karma render provider as well.
thanks for any suggestions
Jeff
My scenes use Karma and I have a single FX license which I believe provides up to 3 Karma render nodes.
I don't mind doing command line renders on each machine if that is what it takes.
What is my best and most cost effective options to employ these machines for rendering my scenes?
Would I need an Engine License for each machine that I need to do a command line render on? $749 per machine?
I am familiar with generating ifd's for Mantra renders and it seems that generating the .usd for Karma is a similar workflow and I am good with that.
My last options seems to be to ditch these machines and turn to a cloud rendering service. Open for suggestions on good Karma render provider as well.
thanks for any suggestions
Jeff
Edited by jeffMc - Feb. 21, 2024 11:17:52
Technical Discussion » Hqueue Server and opening ports
- jeffMc
- 125 posts
- Offline
I finally got this to work by getting my admin to explicitly open the ports just like in all the directions.
Technical Discussion » Hqueue Server and opening ports
- jeffMc
- 125 posts
- Offline
Been digging and I found this command -> netstat -ano | finder -i SYN_SENT
Which when run on the client I believe reports that port 5000 is still being blocked?
TCP 10.x.x.x:59695 192.x.x.x:5000 SYN_SENT 6544
The server is not consistently reporting any blocked ports. If it does it is port :443
Which when run on the client I believe reports that port 5000 is still being blocked?
TCP 10.x.x.x:59695 192.x.x.x:5000 SYN_SENT 6544
The server is not consistently reporting any blocked ports. If it does it is port :443
Technical Discussion » Hqueue Server and opening ports
- jeffMc
- 125 posts
- Offline
I am trying to set up a couple of machines for render nodes in a corporate network - ie lots of security restrictions. My IT wont allow me to directly open ports 5000 and 5001 per all of the Hqueue setup directions I have seen. Instead they are insisting that allowing the apps (client and server) to communicate through Windows Firewall is the same thing. So we made sure to allow Hqserver and Hqclientservice to have access to private networks. Well I am still having trouble getting the clients to see the server and my IT help wants to give up like I dont know what I am doing. I am able to ping the client from the server but the client is unable to ping the server. I also have client running on my server machine and it is able to see itself. Would anyone have any suggestions for a knowledgeable reply to my IT team as to where to look next? Personally I am questioning the equivalence of opening ports vs allowing the app through the firewall but my networking knowledge is pretty limited. Any suggestions much appreciated. Jeff
Solaris and Karma » Depth fading shader
- jeffMc
- 125 posts
- Offline
I am thinking I may need to limit ray refract depth somehow but only on the final image?
Solaris and Karma » Depth fading shader
- jeffMc
- 125 posts
- Offline
Hi I am trying to find a way to make an opacity depth fade for my scene. My output is going to print and I need a way to provide the designers an easy way to fade the deeper parts of the image into a background color they provide.
- A depth AOV is great except it leaves a halo effect from the antialiasing.
- using rayimport (ray:hitdist) on the opacity of my Karma shaders gets close but I don't get a consistent opacity across different objects (see attachment).
Ultimately I would like to be able to avoid post processing an alpha into these images so I think applying a global distance fade seems the best idea. Would anyone have any suggestions for the best approach to this?
thanks
Jeff
ray_through.png
- A depth AOV is great except it leaves a halo effect from the antialiasing.
- using rayimport (ray:hitdist) on the opacity of my Karma shaders gets close but I don't get a consistent opacity across different objects (see attachment).
Ultimately I would like to be able to avoid post processing an alpha into these images so I think applying a global distance fade seems the best idea. Would anyone have any suggestions for the best approach to this?
thanks
Jeff
Image Not Found
ray_through.png
Edited by jeffMc - Dec. 15, 2023 19:01:04
Houdini Lounge » Sticky background image
- jeffMc
- 125 posts
- Offline
Technical Discussion » quick and dirty wireframe preview - how?
- jeffMc
- 125 posts
- Offline
flipbook, aha! yes that is it! exactly what I am after, thank you!
http://www.sidefx.com/docs/houdini/render/flipbook.html [www.sidefx.com]
lol sometimes the things right under my nose are the most difficult to find
http://www.sidefx.com/docs/houdini/render/flipbook.html [www.sidefx.com]
lol sometimes the things right under my nose are the most difficult to find
Technical Discussion » quick and dirty wireframe preview - how?
- jeffMc
- 125 posts
- Offline
i just want to do a quick wireframe or shaded preview of my animation using the screen display.
is there a way to do this without generating a separate image sequence first?
is there a way to do this without generating a separate image sequence first?
Technical Discussion » Moving a scattered geo to the direction of Normals?
- jeffMc
- 125 posts
- Offline
you don' t seem to have any simulation happening here that I can tell so no way to accumulate any movement on the points. @v = @N and so it is everafter.
However you may be looking in the correct direction. There are probably a gazillion ways to do this.
Hope this may point you toward some non simulated options. You could also take this into a solver perhaps or do in a POP for time based action by adding fractions of @N to the points.
at some point, either with simulation or without you need to add the normal to the point position.
this example uses a mix node and attribute transfer to get the normal data from the poly to the point.
hope it is helpful
jammer
However you may be looking in the correct direction. There are probably a gazillion ways to do this.
Hope this may point you toward some non simulated options. You could also take this into a solver perhaps or do in a POP for time based action by adding fractions of @N to the points.
at some point, either with simulation or without you need to add the normal to the point position.
this example uses a mix node and attribute transfer to get the normal data from the poly to the point.
hope it is helpful
jammer
Technical Discussion » mantra_ipr duplicated 700 times
- jeffMc
- 125 posts
- Offline
Happens to me too sometimes. not sure what causes it.
Someone suggested it may be the cat in an earlier thread lol
Someone suggested it may be the cat in an earlier thread lol
Technical Discussion » single thread performance on new Intel processor - turbo boost inverted?
- jeffMc
- 125 posts
- Offline
thanks for this info
I installed HWiNFO and from what I can see I have a theory.
Not being real technical savvy I may be misrepresenting this but it appears that the thread or threads quickly hop from core to core so that the cpu speed I am seeing in the windows task manager is an average of all the cores, not a representation of the speed for the fastest cores.
The individual cores when active do indeed peg out at just under 3.7Ghz but for only a brief moment before maxing out a different core.
The “idle” cores then move to a very low speed which is not surprising.
The more threads working, the more often the individual cores are maxing out and this is reflected in the higher overall speed shown when doing something like a render.
Task manager is showing an average speed for all cores, not the maximum for individual core(s)
So long story short I think all is as it should be.
thanks again
jammer
I installed HWiNFO and from what I can see I have a theory.
Not being real technical savvy I may be misrepresenting this but it appears that the thread or threads quickly hop from core to core so that the cpu speed I am seeing in the windows task manager is an average of all the cores, not a representation of the speed for the fastest cores.
The individual cores when active do indeed peg out at just under 3.7Ghz but for only a brief moment before maxing out a different core.
The “idle” cores then move to a very low speed which is not surprising.
The more threads working, the more often the individual cores are maxing out and this is reflected in the higher overall speed shown when doing something like a render.
Task manager is showing an average speed for all cores, not the maximum for individual core(s)
So long story short I think all is as it should be.
thanks again
jammer
Technical Discussion » single thread performance on new Intel processor - turbo boost inverted?
- jeffMc
- 125 posts
- Offline
I apologize in advance if this seems out of place but I do think it relates to this list as the app I am referring to is Houdini.
I am fortunate enough to have a shiny new dual 8 core Xeon workstation, manufactured by Lenovo. The cpu are the 6134 model with a base frequency of 3.2 and so called turbo mode of 3.7.
My understanding of these processors is that they will increase frequency (turbo boost) when running single threaded apps. However I seem to be getting just the opposite effect on my machine.
When running scenes with (assumed) single threaded routines my cpu clock speed is actually throttled lower, at least according to Task Manager. ie as CPU core utilization decreases so does clock speed.
During some processes the clock will actually decrease to the low 2's Ghz or even lower. It is only when I can load my cores with a multi threaded process or render that I can ever see the speed top 3.5Ghz.
Am I looking at this incorrectly or is there possibly something amiss with my machine?
I have checked the bios and turned off all power saving settings.
thanks in advance for any insight into this
jammer
I am fortunate enough to have a shiny new dual 8 core Xeon workstation, manufactured by Lenovo. The cpu are the 6134 model with a base frequency of 3.2 and so called turbo mode of 3.7.
My understanding of these processors is that they will increase frequency (turbo boost) when running single threaded apps. However I seem to be getting just the opposite effect on my machine.
When running scenes with (assumed) single threaded routines my cpu clock speed is actually throttled lower, at least according to Task Manager. ie as CPU core utilization decreases so does clock speed.
During some processes the clock will actually decrease to the low 2's Ghz or even lower. It is only when I can load my cores with a multi threaded process or render that I can ever see the speed top 3.5Ghz.
Am I looking at this incorrectly or is there possibly something amiss with my machine?
I have checked the bios and turned off all power saving settings.
thanks in advance for any insight into this
jammer
Technical Discussion » Where are help files examples?
- jeffMc
- 125 posts
- Offline
it does! - thank you so much
Obvious but I had no idea that they were available there
jammer
Obvious but I had no idea that they were available there
jammer
Technical Discussion » Where are help files examples?
- jeffMc
- 125 posts
- Offline
I can't seem to connect to help file examples through the internal Houdini ? help.
The pages will display ok but no help files are shown - http://127.0.0.1:48626/nodes/sop/lsystem [127.0.0.1]
If I connect to help through SideFX site the files are there but it says I cant load load them. - http://www.sidefx.com/docs/houdini/nodes/sop/lsystem.html [www.sidefx.com]
I have seen some reference to removing a preferences folder in my $HOME directory but I dont seem to have that folder to start with.
I could really use these example files.
thanks for any suggestions.
jammer
The pages will display ok but no help files are shown - http://127.0.0.1:48626/nodes/sop/lsystem [127.0.0.1]
If I connect to help through SideFX site the files are there but it says I cant load load them. - http://www.sidefx.com/docs/houdini/nodes/sop/lsystem.html [www.sidefx.com]
I have seen some reference to removing a preferences folder in my $HOME directory but I dont seem to have that folder to start with.
I could really use these example files.
thanks for any suggestions.
jammer
Technical Discussion » evolving geometry to wire solver
- jeffMc
- 125 posts
- Offline
I kind of feel I am having a conversation with myself lol.
Thinking this could be helpful to someone else so posting what I think is a reasonable resolution to getting changing geo into and out of the wire solver.
I went down a couple of different paths in trying to do this and it turns out I needed a bit of both.
No effort has been made to style the dynamics here, just glad I was able to get the system working and stable. It is actually kind of weird but I chalk that up to mostly needing to tweak some values (I think).
I have several things in mind for pushing this further. It seems I was stuck at this point for quite a while.
To get it working properly I had to set and trigger a creation frame for each new object as well as have the correct options set on the solver and wire object.
Maybe if I get my momentum back I will post a follow up in the work in progress area.
Pros:
- can add and remove geo at will.
- works reasonably fast (I think)
Cons:
- all geo has to be created in the sop solver. I could not figure out a way to bring in external geo in and have it work. Not a big deal other than having to work in the sop solver.
- could not get a pintoanimation working and did not get around to trying other constraint options.
I would welcome any further suggestions if anyone has any.
jammer
Thinking this could be helpful to someone else so posting what I think is a reasonable resolution to getting changing geo into and out of the wire solver.
I went down a couple of different paths in trying to do this and it turns out I needed a bit of both.
No effort has been made to style the dynamics here, just glad I was able to get the system working and stable. It is actually kind of weird but I chalk that up to mostly needing to tweak some values (I think).
I have several things in mind for pushing this further. It seems I was stuck at this point for quite a while.
To get it working properly I had to set and trigger a creation frame for each new object as well as have the correct options set on the solver and wire object.
Maybe if I get my momentum back I will post a follow up in the work in progress area.
Pros:
- can add and remove geo at will.
- works reasonably fast (I think)
Cons:
- all geo has to be created in the sop solver. I could not figure out a way to bring in external geo in and have it work. Not a big deal other than having to work in the sop solver.
- could not get a pintoanimation working and did not get around to trying other constraint options.
I would welcome any further suggestions if anyone has any.
jammer
Edited by jeffMc - July 23, 2018 19:10:46
-
- Quick Links