michalf
[ Editor ]
We have maya setup this way. While it's super easy to configure in a central place, it does not work without problems. For windows workstations, I had to install maya locally. After installation, you can run server copy of maya and it works most times. We had some minor issues (with muscles IIRC), but this could be due to our messed network setup. We're building new network, so I'll be able to verify this soon. Linux render nodes (which doesn't run GUI) seems to work perfectly.
Nuke and 3delight also run from server copy. I heard one could experience 3delight4maya slow shader translation in case of network install, but I never experienced it. Other minor apps are djv and vDub.
I'd like to know how to install python from central location, especially for windows. Has anyone experience with this ?
EDIT: Recently we tried running more maya instances from network server, and we experienced problems. This was linux64-maya placed on a low performance network server (CentOS/Samba). Trying to run more sessions at the same time put server harddisk under heavy load, and maya refused do start at all. Typical error message was smth like: "Unable to dynamicaly load xxx.so File does not exist" and it was related to maya's native libraries. Moving maya to faster server, or better, copying to local drives solves the problem. So yes, it seems maya doesn't like networked install...
You mean install python remotely or use/execute python over a server on the network?
I wan’t all rendernodes/workstations to share the same python dir, so I can easily add new modules by just placing them in a single dir.For example, instead installing into c:\python2.6, install into \server\Apps\nt\Python2.6
This is basically my question but I can share my thoughts on python: Python is kind of dependent on the registry where it stores the default python installations. This information is e.g. used when you install python package from executable. Other than that the python path and site-packages path must be stored in the PATH environment variable. Both should could be done pointing at the server installation. If UNC paths are a problem network drives should be used. I haven’t tested this though. I thought that if someone had bad experiences I won’t even try. But since there is not much response…
This is basically my question but I can share my thoughts on python: Python is kind of dependent on the registry where it stores the default python installations. This information is e.g. used when you install python package from executable. Other than that the python path and site-packages path must be stored in the PATH environment variable. Both should could be done pointing at the server installation. If UNC paths are a problem network drives should be used. I haven’t tested this though. I thought that if someone had bad experiences I won’t even try. But since there is not much response..
I can confirm that Nuke does work well from a central location. If you have a slow connection to that central location, startup times can be extremely long. But installing once for an entire facility is quite handy.