[ghost82]

Originally Posted by aural

Dear Daniele,
I have two win10 PCs, connected by a router. I set the same user name and password for both PCs, but intel mpi wouldn't run. I wonder if the two PCs have to be in the same domain (they are currently in the same workgroup) to make it work. Thank you for your help,

Try to disable the windows 10 firewalls on both machines, to check "select IP interface" and to select the right ip interface (choose interconects defaults and intel mpi types): till now it's the only method is working for me (however I'm not using a router/switch), let me know.
Yes, set the same workgroup and ping each other to check if they see each other, after disabling the firewalls and before doing anything else.

[ghost82]

Originally Posted by ghost82

Try to disable the windows 10 firewalls on both machines

After everything is ok with the windows firewalls disabled you can try to reactivate them.
After playing a bit with the loggers I was able to re-enable the windows 10 firewalls on both machines.
Here is my setup:


MACHINE 1 (this machine is the host + 16 nodes)
Computer name: workstation
OS: Windows 10 64 bit Professional, with latest updates as of the date of this post
Network: Infinihost III LX PCI Express x8 HCA adapter MHGS18-XTC
Ansys fluent: v.19.2
MPI Platform: Intel MPI (from ansys 19.2 DVD media)



MACHINE 2 (this machine is a 16 nodes)
Computer name: workstation-2
OS: Windows 10 64 bit Professional, with latest updates as of the date of this post
Network: Infinihost III LX PCI Express x8 HCA adapter MHGS18-XTC
Ansys fluent: v.19.2
MPI Platform: Intel MPI (from ansys 19.2 DVD media)


When you start fluent on the host machine the firewall will ask for some processes/programs to access the network, give access to these programs.
In addition I had to enable some additional rules, which were not asked by windows 10, in details:


- on the host machine (machine 1):
go to control panel -> system and security -> windows defender firewall -> advanced settings


Select on the left "incoming connections rules" and click on the right on new rule;
choose program -> next -> program path -> choose ->

\\workstation\ansys inc\v192\fluent\fluent19.2.0\win64\2ddp_node\fl_mpi1920.exe

Give a custom name for this rule


Select on the left "incoming connections rules" and click on the right on new rule;
choose program -> next -> program path -> choose ->

\\workstation\ansys inc\v192\fluent\fluent19.2.0\win64\2d_node\fl_mpi1920.exe

Give a custom name for this rule


Select on the left "incoming connections rules" and click on the right on new rule;
choose program -> next -> program path -> choose ->

 \\workstation\ansys inc\v192\fluent\fluent19.2.0\win64\3ddp_node\fl_mpi1920.exe

Give a custom name for this rule


Select on the left "incoming connections rules" and click on the right on new rule;
choose program -> next -> program path -> choose ->

\\workstation\ansys inc\v192\fluent\fluent19.2.0\win64\3d_node\fl_mpi1920.exe

Give a custom name for this rule


N.B.: change "workstation" with your computer name (host machine)




- on the node machine (machine 2):
go to control panel -> system and security -> windows defender firewall -> advanced settings


Select on the left "incoming connections rules" and click on the right on new rule;
choose program -> next -> program path -> choose ->

%ProgramFiles% (x86)\IntelSWTools\compilers_and_libraries_2017.4.210\windows\mpi\intel64\bin\hydra_service.exe

Give a custom name for this rule



Select on the left "incoming connections rules" and click on the right on new rule;
choose program -> next -> program path -> choose ->

 \\workstation\ansys inc\v192\fluent\fluent19.2.0\win64\2ddp_node\fl_mpi1920.exe

Give a custom name for this rule


Select on the left "incoming connections rules" and click on the right on new rule;
choose program -> next -> program path -> choose ->

\\workstation\ansys inc\v192\fluent\fluent19.2.0\win64\2d_node\fl_mpi1920.exe

Give a custom name for this rule


Select on the left "incoming connections rules" and click on the right on new rule;
choose program -> next -> program path -> choose ->

 \\workstation\ansys inc\v192\fluent\fluent19.2.0\win64\3ddp_node\fl_mpi1920.exe

Give a custom name for this rule


Select on the left "incoming connections rules" and click on the right on new rule;
choose program -> next -> program path -> choose ->

\\workstation\ansys inc\v192\fluent\fluent19.2.0\win64\3d_node\fl_mpi1920.exe

Give a custom name for this rule


N.B.: change "workstation" with your computer name (host machine)




Remember to start fluent with these options:
- Working directory: \\workstation\path-to-fluent-working-directory
- Fluent Root Path: \\workstation\path-to\Ansys Inc\v192\fluent


N.B.: change "workstation" with your computer name (host machine)


Parallel settings:
- Interconnects: default
- MPI Types: intel
- check Select IP interface and choose your network (in my case 192.168.3.1)

----------------------------------------------------------------------



In case of trouble with firewalls you can enable in the windows event logger which program name give you problems (following instructions courtesy of https://resque.wordpress.com/2013/11...-event-access/ ):


1. type "gpedit.msc" in start > run (or the command prompt) to open the group policy mmc snap in editor

2. Browse to the sub folder
Local Computer Policy > Computer Configuration > Windows Settings > Security Settings > Local Policies > Audit Policy

3. right click on "Audit object access" and click properties

4. select the check box next to the word "failure" on the first tab marked "Local Security Settings"

Now when ever windows firewall blocks an application from accessing a certain port you will be able to see the loged even in the windows event log under the security sub folder

To access the event log open

1. Control Panel > Administrative Tools > Computer Management
2. Then under the sub folder "Computer Management (Local)" > Event Viewer > Windows Log > Security

[aural]

Originally Posted by ghost82

Try to disable the windows 10 firewalls on both machines, to check "select IP interface" and to select the right ip interface (choose interconects defaults and intel mpi types): till now it's the only method is working for me (however I'm not using a router/switch), let me know.
Yes, set the same workgroup and ping each other to check if they see each other, after disabling the firewalls and before doing anything else.

Thank you Daniele.
I disabled firewalls and defender, it couldn't run if I select Intel as the MPI. However, the program was launched successfully if I select IBM as the MPI.

The difference between these two is Intel requires to register both username and password, but IBM requires to register password only. So I wonder if the unsuccess by using Intel is I missed something in this step? For user name input, should I use "computer name/username" format or just "username" will be sufficient? Thank you,

[LuckyTran]

Originally Posted by aural

For user name input, should I use "computer name/username" format or just "username" will be sufficient? Thank you,

Just username should be sufficient because the mpi wants to launch the process on all the machines using the same username. The computername/username format allows you to log onto another machine using a username that exists only on that machine and not your own. As far as I know, mpi's don't work this way.

[aural]

Originally Posted by LuckyTran

Just username should be sufficient because the mpi wants to launch the process on all the machines using the same username. The computername/username format allows you to log onto another machine using a username that exists only on that machine and not your own. As far as I know, mpi's don't work this way.

Got you, Thanks!

[ghost82]

ok...time to ask help to you now
I just updated to the latest release of fluent, 2019 R1 on win10.
Serial works fine on both machines.
Parallel works fine on both machines (individually).
I'm using as usual intel mpi.
I registered, as usual domain/user and password on both machines.

Parallel with 2 machines doesn't work, the gui stucks at "Host spawning Node 0 on machine "Workstation" (win64)".

No error after that line, I have a log file in my working directory (debug.log), which I never saw before, with this:
[some numbers]WARNING:resource_bundle_qt.cpp(115) locale_file_path.empty() for locale

Firewalls are disabled, I don't know where to start to debug the problem...

Anybody able to run parallel on multiple machines with 2019 R1 and win10?

[LuckyTran]

Originally Posted by ghost82

No error after that line, I have a log file in my working directory (debug.log), which I never saw before, with this:
[some numbers]WARNING:resource_bundle_qt.cpp(115) locale_file_path.empty() for locale

The warning smells like notepad is involved.

I have heard there is a hidden file called debug normally associated with notepad in C:\Users\(User-Name)\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\Startup

go there and delete it. But why it appears in your working directory? Something must be creating it.

[ghost82]

Originally Posted by LuckyTran

The warning smells like notepad is involved.

I have heard there is a hidden file called debug normally associated with notepad in C:\Users\(User-Name)\AppData\Roaming\Microsoft\Windows\Start Menu\Programs\Startup

go there and delete it. But why it appears in your working directory? Something must be creating it.

Thank you for yur reply, but no hidden file in that folder.
I opened a ticket with ansys support, smells like a bug to me...I will update the post.
The debug file is created only when parallel is setup to be between two pcs, nothing is created and all is working ok with intel mpi on each workstation.

[LuckyTran]

So I learned that debug.log is created windows (or maybe visual studio) and gets placed in the same directory as the executable. My particular example was just one dealing with autostarting of notepad that wasn't autostarting.

resource_bundle_qt.cpp(115) locale_file_path.empty() for locale means qt couldn't find resource bundle. This usually means a build error which normal people would fix by uninstalling and reinstalling. But your Fluent is running fine locally on each machine, so it's probably worth a ticket.

Since it's a build error, it would be really hard to debug. It's possible that only you have this issue and no one else. Or the next two machines that you install Fluent on, it might suddenly work.

Btw, did 19.0 work okay?

[ghost82]

I'm sure 19.2 was ok, it was the first version running in my windows 10.


Edit: the same behaviour happens with IBM mpi

[ghost82]

Originally Posted by ghost82

I just updated to the latest release of fluent, 2019 R1 on win10.
Serial works fine on both machines.
Parallel works fine on both machines (individually).
I'm using as usual intel mpi.
I registered, as usual domain/user and password on both machines.

Parallel with 2 machines doesn't work, the gui stucks at "Host spawning Node 0 on machine "Workstation" (win64)".

Problem was solved thanks to ansys support.


The story
------------


first ticket at ansys support was opened on January 14th.
Unfortunately the engineer who was assigned my ticket wasn't able to solve the issue and he was requesting a webex chat: I was out of office in that days and without my reply the ticket was closed....I was not very happy....


Second ticket was opened last sunday and the problem was solved just now by engineer Mark Owens and the developer team.
My congratulations to him, after few days of tests and debugs he was able to spot the problem and give us a simple solution.


The problem
--------------
After installing Ansys 2019 R1 on 2 workstations with windows 10 professional build 1809 and after installing the intel mpi package provided by default by ansys dvd media, fluent parallel computations weren't working: fluent was hanging at "Host spawning Node 0 on machine "Workstation" (win64)".
By analyzing the behaviour of the mpiexec process it seemed that the first machine was trying to send a tcp packet without any reply from machine 2.
Ansys cfx was working without problems, so the errors were confined to fluent.


The solution
---------------
In a few words, the problem was with the mpi package: if you install the intel mpi package from the ansys installation window, the default version is 2017.3.210.
This version generates these errors; solution is to install a different version, in particular 2018.3.210.
Version 2018.3.210 can be found in X:\Program Files\ANSYS Inc\v193\commonfiles\MPI\Intel\2018.3.210\Windows\ w_mpi_p_2018.3.210.exe after you install ansys on your pc.
So uninstall any version of intel mpi and install v. 2018.3.210 on all workstations.


Sidenotes
----------------
At the beginning, we were thinking that the intel mpi package should be the cause of our problems, so we downloaded a newer version of intel mpi, more into details v. 2019.1.144.
We had no luck, errors were generated with this newer version: we have already informed ansys about this problem for future releases.


Apparently, the debug.log file created in the working directory, which has inside the line "resource_bundle_qt.cpp(115) locale_file_path.empty() for locale" is not related at all with mpi and it's a different topic, maybe related to a bug with Qt. It's a warning, not an error, so this shouldn't hurt fluent.


Thanks
------------
Again, our thanks go to Mark Owens and to the developer team, great professional support.


Hope this topic will help someone else.



PS: always have a look at the system requirements for mpi: as you can see in the attachment, the correct version for intel mpi is 2018.3.210, so here the information is correct for intel mpi.
However, the problem here is that v. 2017.3.210 is installed from the installation gui, which is not compatible.

[ghost82]

Edit: Ansys 2019 R2 compatible with intel mpi 2018.3.210

[Roil]

I have this problem. Is there anyone here who can help me solve this problem?

Originally Posted by xh110120

Hi,

I have focused on the muti-machine parallel for fluent 14.5 these days. And there is always a problem. The relative works have been done. But after opening the fluent in the host computer, the state will always get stuck at this step"checking the status of SMPD for INTEL MPI on the local machine...smpd runing on tan-PC". And I also check the task manager seperately in the host computer and node computer. The corresponding cores have been working, and the process of smpd also appears in the task manager. I don't know what problems it has. And I've tried every method I could think, while they don't work. Could you please help me to find another way to figure it out? Thank you very much for your kindly help!

[elochana]

Hello,

Thank you very much for this tutorial. I am trying to get this done with 2 laptops in Ansys 20R2. I also followed your updates. But, as you experienced before changing to mpi 2018.3.210, I get the same error even after changing to mpi 2018.3.210 which is included in Ansys 20R2. My fluent instance gets stuck at "Host spawning Node 0 on machine ... "

After sometime, it exits and say "The Fluent application failed to validate the connection.".

Where exactly should I type "wmpiregister". When I do that cmd says "unrecognised...." After installing mpi 2018.3.210 a Getting started html is opened in browser which asks to run "> mpiexec -register". I think it does the same thing as "wmpiregister". So I did that. The html also asks to run "> hydra_service -install" and "> hydra_service -start". I did them too.

Any advice would be greatly appreciated.

Thank you

[Hasan_Najafi]

Originally Posted by ghost82

Edit: Ansys 2019 R2 compatible with intel mpi 2018.3.210

Hi
Daniele
I have a problem with parallel processing with two pcs. When I am running the Fluent I see this window. After I put the user name. I can't write the password. Is similar to your previous experience.

Thanks