issue with getFieldURI in distributed setting

[grauchs]

I manage to run my mupif script on one machine, using two server applications on a different machine, transgferring the fields by getField and setField methods. If I replace the getField method by getFieldURI, the execution hangs in the subsequently called setField method. However, if I run the control script and the two server applications on the same machine, it executes without any problem until the end.
the relevant part in the control script is:

---

print('$$$ Displacement output')
uri = app1.getFieldURI(FieldID.FID_Displacement,istep.getTime().inUnitsOf('s').getValue())
log.info("URI of problem 1's Displacement field is " + str(uri) )
res1 = Pyro4.Proxy(uri)
#res1 = app1.getField(FieldID.FID_Displacement,istep.getTime().inUnitsOf('s').getValue())
res1.toVTK2('testoutput-d',format='ascii')

print('******** SET FIELD TO SECOND MODEL ')
print('*************************************')
print('$$$ Field 1 time',res1.getTime())
print('$$$ set displacement field to model 2')
app2.setField(res1)

---

I noticed that I can address field res1 in the controlscript, as in print('$$$ Field 1 time',res1.getTime()), but not in the subsequent setField executed in the API. I noticed that in the API's setField method, the field is treated as an Pyro4.core.Proxy, and subsdequent Field methods do not execute. As mentionned above , the problem does not show up if the control script executes on the same machine as the two application servers!

For information, here are the relevant parts Config.py and two Serverconfig.py Files

---

        self.nshost = '10.1.1.231'
        self.nsport = 9090

        self.server = '10.1.1.232'
        self.serverPort = 44382
        self.serverNathost = '127.0.0.1'
        self.serverNatport = 5555

        self.server2 = '10.1.1.232'
        self.serverPort2 = 44385
        self.serverNathost2 = self.server2
        self.serverNatport2 = 5558
        self.appName2 = 'MuPIFServer2'

---

    self.serverPort = self.serverPort+1
    if self.serverNatport != None:
        self.serverNatport+=1
    self.socketApps = self.socketApps+1
    self.portsForJobs=( 9200, 9249 )
    self.jobNatPorts = [None] if self.jobNatPorts[0]==None else list(range(6200, 6249)) 

---

    self.serverPort = self.serverPort+2
    if self.serverNatport != None:
        self.serverNatport+=2
    self.socketApps = self.socketApps+2
    self.portsForJobs=( 9250, 9300 )
    self.jobNatPorts = [None] if self.jobNatPorts[0]==None else list(range(6250, 6300)) 

---

[nitramkaroh]

We tried the Example11, where the getFieldURI method is used, and it is working in our setting in Linux as well as in Windows.
Do you use vpn or ssh?
Can you check that your ports are open?

[grauchs]

I use ssh.
On the application server, running the two applications, the following relevant ports are open.
Proto Recv-Q Send-Q Local Address Foreign Address State
tcp 0 0 10.1.1.232:9250 0.0.0.0:* LISTEN
tcp 0 0 127.0.0.1:6250 0.0.0.0:* LISTEN
tcp 0 0 10.1.1.232:9200 0.0.0.0:* LISTEN
tcp 0 0 127.0.0.1:10001 0.0.0.0:* LISTEN
tcp 0 0 127.0.0.1:10002 0.0.0.0:* LISTEN
tcp 0 0 10.1.1.232:44382 0.0.0.0:* LISTEN
tcp 0 0 10.1.1.232:44383 0.0.0.0:* LISTEN
tcp 0 0 ::1:6250 :::* LISTEN

On the machine running the control script, the following ports are open
Proto Recv-Q Send-Q Local Address Foreign Address State
tcp 0 0 127.0.0.1:6250 0.0.0.0:* LISTEN
tcp 0 0 127.0.0.1:5556 0.0.0.0:* LISTEN
tcp 0 0 127.0.0.1:5557 0.0.0.0:* LISTEN
tcp 0 0 127.0.0.1:6200 0.0.0.0:* LISTEN
tcp 0 0 ::1:6250 :::* LISTEN
tcp 0 0 ::1:5556 :::* LISTEN
tcp 0 0 ::1:5557 :::* LISTEN
tcp 0 0 ::1:6200 :::* LISTEN

So basically, one the application server, port 6200 is not open, despite appering in the log of the server. I don't know if this is critical.

[grauchs]

I managed to reproduce the error with the demoapp! I put it into a tarball. To run it, you have to set the ip of the nameserver and set some paths in th Config.py file, then start Server1.py in the ServerA directory, then start Server2.py in ServerB, the Example666.py in HQ, the last starting the computations. There is a flag in Example666.py for switching between getField and getFieldURI. If Server1.py runs on a different machine that Example666.py, then the application will hang when it sets the field. Use the -m1 flag for ssh communication.
testclone.zip
If there are problem running the example, don't hesitate to ask

[nitramkaroh]

Dear Gaston,

we hope that we finally solved the issue. Can you, please, try to run the same example using the latest mupif git version?

Thank you.
Martin