DR/networking tips

[Frood]

I know that I should put all textures and assets in one place in the network but for now I don't have a server I only have 2 computers I use one as a WS and the other one as render node it is very annoying to copy all assets each time between 2 pcs (...)

There is absolutely no need for a (dedicated/real) server if you are working with just a few computers. Just create a network share with proper rights on one of your machines and use it as the asset resource for all participating nodes (using UNC paths of course).

Good Luck

[antanas]

@Bormax @PROH - well that explains why I haven't noticed that behavior - it seems I always or at least mostly use corona's stop button for that.

@mahmoudkeshta @Frood - I use simple mapped network drives for that, well as an example on my main workstation - drive D holds textures and models, drive E holds currently worked on projects, and those two are mapped to be the same across all of my home network, well to be the same D and E on all the pc's I mean, users and passwords are the same on all of those PC's too.
 Not a single problem with that method through the years of usage and there's no need to use UNC patch names which is imo way faster to work with. The only problem with that method is of course the network and hdd speed limitations and even if gigabit network is mostly sufficient for most of the cases hdd speed, most often is not but that's another story and the same limitations apply when unc patch method is used too.
 The fastest and most reliable method would perhaps be the automatically cloned\synchronized drives across those network pc's as astudio suggested in this https://forum.corona-renderer.com/index.php/topic,2340.msg17492.html#msg17492 thread but for me it is not currently an option as that means buying additional 10 4tb drives which would cost quite a sum and would considerably hit my budget, but yeah, I guess the render start speed improvements using DR with some of the heavier scenes would be tremendous with that method not to mention automatic backups etc.     

[mahmoudkeshta]

There is absolutely no need for a (dedicated/real) server if you are working with just a few computers. Just create a network share with proper rights on one of your machines and use it as the asset resource for all participating nodes (using UNC paths of course).

Good Luck

@mahmoudkeshta @Frood - I use simple mapped network drives for that, well as an example on my main workstation - drive D holds textures and models, drive E holds currently worked on projects, and those two are mapped to be the same across all of my home network, well to be the same D and E on all the pc's I mean, users and passwords are the same on all of those PC's too.
 Not a single problem with that method through the years of usage and there's no need to use UNC patch names which is imo way faster to work with. The only problem with that method is of course the network and hdd speed limitations and even if gigabit network is mostly sufficient for most of the cases hdd speed, most often is not but that's another story and the same limitations apply when unc patch method is used too.
 The fastest and most reliable method would perhaps be the automatically cloned\synchronized drives across those network pc's as astudio suggested in this https://forum.corona-renderer.com/index.php/topic,2340.msg17492.html#msg17492 thread but for me it is not currently an option as that means buying additional 10 4tb drives which would cost quite a sum and would considerably hit my budget, but yeah, I guess the render start speed improvements using DR with some of the heavier scenes would be tremendous with that method not to mention automatic backups etc.     

Thanks for your suggestions I will try them.

@antanas - What if I have the same drives letters on the render node ? will it cause a conflict in paths ?

[antanas]

Well of course there will be conflicts if same named drives are already present but more specifically, you wouldn't be able to map those at all using the same names - in my case all other physical drives present on rendernodes got different names, well, drives D or E or F or G are physically present only on my main workstation but are mapped as such across all my other pc's even notebook - all other physical rendernode pc drives are named accordingly without name overlapping\duplication across the network and are mapped on main workstation as well http://c2n.me/3xmSWj1 - that's how it looks on my main workstation and this http://clip2net.com/s/3xmUaeg is how it looks on one of the secondary/rendernode ones - and yes it can be messy but is the easiest way to manage all that data, asset/texture/proxy/xref paths and so on or at least for me it is the easiest and most understandable way to do that ))

[mahmoudkeshta]

Well of course there will be conflicts if same named drives are already present but more specifically, you wouldn't be able to map those at all using the same names - in my case all other physical drives present on rendernodes got different names, well, drives D or E or F or G are physically present only on my main workstation but are mapped as such across all my other pc's even notebook - all other physical rendernode pc drives are named accordingly without name overlapping\duplication across the network and are mapped on main workstation as well http://c2n.me/3xmSWj1 - that's how it looks on my main workstation and this http://clip2net.com/s/3xmUaeg is how it looks on one of the secondary/rendernode ones - and yes it can be messy but is the easiest way to manage all that data, asset/texture/proxy/xref paths and so on or at least for me it is the easiest and most understandable way to do that ))

I got it .. you mean that I should use a unique letters for work drives through the network then map them to be the same names and paths on all computers .. I am a bit beginner in dealing with network but it is clear now ... as you mentioned before the main problem will be on the network speed and it may affect on reading the assets specially in complex scenes with big size assets ... thanks a lot for your help :)

[FrostKiwi]

I got it .. you mean that I should use a unique letters for work drives through the network then map them to be the same names and paths on all computers

Naaaah.
You really should not assign Drive letters or mount drives as network drives from another PC.
You really really should use complete paths.
They should look like this:
//PC-with-render-data/Shared-network-directory/texture-folder/texture.jpg
Not like this: X:/texture-folder/texture.jpg

If you assign letters, these letter are only known locally and can and will change and will screw everything up. One of these trivial things, that you are sure are too simple to fail and then you search why DR does not find assets and everything is sad, it rains and the coffee is cold :[

[Juraj]

I've been using mapped drives for 4 years, with up to 10 devices, never had a single issue. (OK, single yes, but that's Windows and its weird elevation rights, can be solved in 2 clicks in register) Don't create false panic ;- )

UNC paths are all cool and dandy, but unless you work in environment where it's necessity, you don't need to bother. Mapped drives are completely fine to use. More user-friendly and faster to use for non-technical people, and actually easier transferable because they're uncoupled from physical location of files. I can transfer my harddrives between NAS and fileserver without touching paths.

On other hand, mapping system drive and giving unique letters to system drives of other devices is definitely not how it's supposed to be done :- D

[antanas]

TOTALLY agree with that - not a single issue for almost 10 years of using that method or at least as far as I can remember using DR and there are 7 pc's in my home network 3 of them are workstations and 4 are rendernode only - not too many but still.

"On other hand, mapping system drive and giving unique letters to system drives of other devices is definitely not how it's supposed to be done :- D " - well, If you meant the ugly drive letter chaos on my screenshots then, perhaps )) but for me it is quite handy (think of a storage cluster or some sort of a nas storage)) and besides I do not map system drives (but I would if I could)), only secondary\storage ones, well, for obvious reason - ease of use of that, otherwise quite hard to get to rendernode\secondary workstation storage space from my main workstation and vice versa )

[mahmoudkeshta]

There is misunderstand here perhaps from my side because I am not familiar with network issues :) .

To be more clear I have 2 computers one is the main workstation and the other one is render node. On the workstation I have all my work files and assets on (E) Drive. I won't change anything on the workstation drive letters but I will create mapped drive on the render node with E letter directed to the E drive on the workstation to use the same paths ( actually I need to change the original letters of the render node to another letters because now it has letter E also ). Is this correct If I will use the mapped drive method  ?

[antanas]

Yep, that's basically it, just don't forget to setup user rights (I create the same user with the same password on all my pc's) etc. for that rendernode to be able to access that workstation's (and now it's too) E drive and then starting to work, start workstation pc first and start it some minutes prior to starting rendernode so ws's windows fully loads + loads all it's programs firewalls etc. and then start rendernode so it can refresh that mapping automatically otherwise it can leave it disconnected until you manually access that drive from that rendernode - maybe there's is some program or some automatically launched command script or even some windows function that could do that accessing\refreshing automatically but I didn't come across such a thing, not that I searched too hard though ).

[Juraj]

just don't forget to setup user rights (I create the same user with the same password on all my pc's)

You can just create 'Workgroup'. This way any computer can access others in network without accounts.

[Nekrobul]

IDK if this will help anyone but my workflow with DR goes like this.

- Setup pasword and login on all PC's wich sould be linked in DR
- Setup network drive on the leadig PC and Slave PC's (Using UNC for destination on the network drive //192.168.bla.bla/blablabla not //blablabla/blablabla)
- Making shure that the instalation (MAX,plugins etc) is the same on all nodes including master PC.
- Making an achive of the scene saving it directly in to shared folder linked to the netwok drive.
- Transfering all data linged to the scene (textures,proxies,simfiles etc) in to one folder called MAPS.
- Opening scene trough the network drive on the mastr PC.
- Trough the asset tracking rewrinting paths of the all linked files to the network drive and the folder we just created.
- Converting them to UNC location.
- Serching lan for nodes.
- Setuping update interval for 3sec and max pixel transfer to 100000
- Presing render.
- ???
- Profit.

PS - for comfortable use of MANY nodes it is better to use some remote connection software like mRemoteNG for example.

[Frood]

I didn´t wanted to reply to any post here (even haven´t started this thread, also if it says so because maru thought to remove the spam from the daily builds thread which was good) but this is a very ... errr.. complicated approach to use DR rendering :) Even if using mapped drives, the workflow should be (and can be):

- Create scene
- Activate DR
- Render :)

Good Luck

[Juraj]

I didn´t wanted to reply to any post here (even haven´t started this thread, also if it says so because maru thought to remove the spam from the daily builds thread which was good) but this is a very ... errr.. complicated approach to use DR rendering :) Even if using mapped drives, the workflow should be (and can be):

- Create scene
- Activate DR
- Render :)

Good Luck

That's exactly how it works in my case. With Vray, Maxwell...Corona. I never did anything else than.... click create Workgroup (Homegroup back in WindowsXP) on any PC, click join on all other PCs. Click mapped drive from fileserver (could be master workstation, doesn't matter), voila. Done.

The issue doesn't seem to be about conflict between mapped drives and UNC paths, but rather than most users still use some locally stored files with either direct paths or just searchable external paths. (like big 1TB drive that hosts both system and assets on master PC). You could just convert such files to UNC directly.

[mahmoudkeshta]

I didn´t wanted to reply to any post here (even haven´t started this thread, also if it says so because maru thought to remove the spam from the daily builds thread which was good) but this is a very ... errr.. complicated approach to use DR rendering :) Even if using mapped drives, the workflow should be (and can be):

- Create scene
- Activate DR
- Render :)

Good Luck

Totally agree. Until now I can't do it in a right way with mapped drive. also I don't know how to convert to UNC paths in asset tracking I tried every single option in it but no way.

My current problem is the render node is reading network path while the master pc is reading local paths. 

" Transfer missing assets " in Vray is like a magic.

That's exactly how it works in my case. With Vray, Maxwell...Corona. I never did anything else than.... click create Workgroup (Homegroup back in WindowsXP) on any PC, click join on all other PCs. Click mapped drive from fileserver (could be master workstation, doesn't matter), voila. Done.

The issue doesn't seem to be about conflict between mapped drives and UNC paths, but rather than most users still use some locally stored files with either direct paths or just searchable external paths. (like big 1TB drive that hosts both system and assets on master PC). You could just convert such files to UNC directly.

How can I convert paths to UNC from old files? or I should work with them from beginning ?