HQueue Path mapping in mixed env and timeout

[uzer_name]

Hi All

I have couple small problems in tuning HQueue farm

1. workstations Linux and win based
2. farm - mixed (osx/linux/win)
hq_shared storage  - nfs/smb NAS

OSX and linux configured with autofs service, so Projects located on nfs share looks like "/net/servername/raid/nfs_sharename/Projects/XXX" for both osx and linux rendernodes and workstations. So i could render/simulate/whatever on linux/osx nodes with current project file, no need to copy hip to "Common Shared Storage". Just use "render current hip" in Hq_render node.

But when i try to use win nodes and win workstation - arising problems.
shared project for win part look as "//servername/raid.sharename/Projects/XXX" and "Projects" is a smb share mounted as disk "P:/" on nodes and worksations

so when i try to send hip to render from linux/osx workstation usin "render current hip" to win nodes, nodes shows error "cannot read hip file "/net/servername/raid/nfs_sharename/Projects/XXX/scenes/file.hip" - seems Hqserver do not translate linux/osx path to win path. Same when using "render target hip" .  But when i use "copy to shared storage" and HQ_render node copies hip and all needed cache/img files to "$HQROOT/projects" - win stations renders ok. But cannot do distributed simulation (error with saving cache file to *nix path)

is it possible to use win nodes for render/simulate hip from osx/linux workstations as "current hip" , assuming that hip located on NAS, accesible from all stations/nodes?

how to correctly set HOUDINI_PATHMAP var? Because in help/faq exist different var format examples like
a. HOUDINI_PATHMAP = "/unix/path1" "//win/path1", "/unix/path2" "//win/path2"
b. HOUDINI_PATHMAP = {"/unix/path1":"//win/path1", "/unix/path2":"//win/path2"}
c. HOUDINI_PATHMAP = "/unix/path1":"//win/path1"; "/unix/path2":"//win/path2"
etc.
and no luck woth using any in houdini.env - all gives format/var error.

and 2nd problem

When i render on linux/osx nodes, and rendertime for 1 frame quite small (1-2min), linux and osx nodes finishing task, write image, and stays stuck on 100% completed task and waiting 6-8 minutes.
in hqueue client page of those stuck stations shows quite high idle numbers (0.5 ... 2), so i think hqueue assumes that those stations in busy state (in fact no - 0% cpu usage by user task). is it any timeout/idle calculation settings/limits in hqueue settings to cure such timeouts?

Thanks in advance.

 

Edited April 9, 2017 by uzer_name