pconnect function returns handshake error

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

pconnect function returns handshake error

Prasad K
Hi all,
We are trying to run octave 4.0.0 with multiple nodes (each node consisting
of 16-cores) using cluster functions in parallel toolbox v3.1.1.
Below is the error we are getting when we try to connect to server as said
in  this link
<https://octave.sourceforge.io/parallel/package_doc/Example.html>  

"octave:2> conns = pconnect("server")
error: handshake failed
error: could not create command stream to server
error: can't poll since command stream not good
error: could not close command stream"




--
Sent from: http://octave.1599824.n4.nabble.com/Octave-General-f1599825.html


Reply | Threaded
Open this post in threaded view
|

Re: pconnect function returns handshake error

Olaf Till-2
On Tue, May 22, 2018 at 05:57:34AM -0700, Prasad K wrote:
> Hi all,
> We are trying to run octave 4.0.0 with multiple nodes (each node consisting
> of 16-cores) using cluster functions in parallel toolbox v3.1.1.
> Below is the error we are getting when we try to connect to server as said
> in  this link
> <https://octave.sourceforge.io/parallel/package_doc/Example.html>  
>
> octave:2> conns = pconnect("server")

That's not in the example, and not supposed to work, it should be:

conns = pconnect ({"server"})

But maybe you actually have done this?

Olaf

--
public key id EAFE0591, e.g. on x-hkp://pool.sks-keyservers.net



signature.asc (849 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: pconnect function returns handshake error

Prasad K
Thanks Olaf

Yes, we have done that to be exact but it gives the same error.

"octave:3> conns = pconnect({"server"})
error: handshake failed
error: could not create command stream to server
error: can't poll since command stream not good
error: could not close command stream"




--
Sent from: http://octave.1599824.n4.nabble.com/Octave-General-f1599825.html


Reply | Threaded
Open this post in threaded view
|

Re: pconnect function returns handshake error

Olaf Till-2
On Tue, May 22, 2018 at 09:30:18PM -0700, Prasad K wrote:
> Thanks Olaf
>
> Yes, we have done that to be exact but it gives the same error.
>
> "octave:3> conns = pconnect({"server"})
> error: handshake failed
> error: could not create command stream to server
> error: can't poll since command stream not good
> error: could not close command stream"

There was a bug in parallel-3.1.1 which could produce such an error
(maybe depending on the compiler(-version) used to compile the
package) if TLS is used (but it may be dangerous _not_ to use
TLS). This bug is fixed in the current version of parallel (3.1.2).

There could also be a configuration error. (Same version of parallel
must be at all machines, TLS must be used at all machines or at none,
TLS authentication files must be at correct positions at all
machines.)

The server has logfiles (more then one): /tmp/octave_error-.....log
which could provide further information.

If you have no success after upgrading to parallel-3.1.2, you should
submit a bug report. In this case, the logfiles of the server should
also be provided, and it could help to uncomment the lines:

// #define octave_parallel_debug_server 1
// #define octave_parallel_debug_client 1
// #define octave_parallel_debug_lib 1

in src/parallel-gnutls.h before compiling the package -- but take
care, with this lines uncommented, sensitive data will be written to
standard error and to the logfiles. (This means, among other things,
that you should mask these data before sending the files to the
mailing list.)

Olaf

--
public key id EAFE0591, e.g. on x-hkp://pool.sks-keyservers.net



signature.asc (849 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: pconnect function returns handshake error

Prasad K
Thanks Olaf

Upgrading the package parallel form v3.1.1 to v3.1.2 did help to overcome
that error and successful connection is established between the client and
server. But parallel package v3.1.2 was  giving errors when trying to build
on octave v4.0.0 so i also upgraded octave to v4.4.0.



--
Sent from: http://octave.1599824.n4.nabble.com/Octave-General-f1599825.html