MathGroup Archive 2005

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: Re: Lost connection to license server

  • To: mathgroup at
  • Subject: [mg62439] Re: [mg62407] Re: [mg62283] Lost connection to license server
  • From: "Yasvir A. Tesiram" <tesiramy at>
  • Date: Thu, 24 Nov 2005 06:33:38 -0500 (EST)
  • References: <> <Pine.LNX.4.63.0511201648460.10283@localhost.localdomain> <> <> <>
  • Sender: owner-wri-mathgroup at

OK, I think I understand. I just simulated what you've described, one 
level deep only. I launched a remote kernel from my notebook 
running Linux. The remote kernel was MacOS X. I launched a kernel 
with LinkLaunch on the same machine. So now I have two MathKernels.
I killed the license server. After a long time messages started appearing 
and all evaluated fine after that. 
Without any code on how you launched your remote sessions, I don't want to 
spend the time to simulate connections N levels deep. Sorry. But here are 
some notes and suggestions.

When a the math kernel is started as far as I can tell by monitoring it 
with top (in Unix), the kernel keeps running even if MathLM crashes. So if 
all your remote launches lead to one interface (either the notebook or the 
command line session) from your machine, evaluating from that one 
interface after you restart MathLM should get you back on track with all 
other machines.

After LinkLaunch you may want to add in a LinkReadyQ to test 
whether the various links are ready or not and you may want to keep trying 
until it returns true.

You may want to look at gridMathematica for your needs, particularly, with 
20 machines doing "things" for you.


On Wed, 23 Nov 2005, Manuel Schmidt wrote:

> Hi,
> my problem is exactly, that the message "...could not contact..." does
> NOT pop up. It will never pop up, because the Kernel is started by
> another Kernel and not by the FrontEnd.
> If I click on "retry" butto of the kernel, that has been started by the
> FrontEnd, everything is ok. But from the Kernel started by the
> FrontEnd's Kernel I (i.e. FrontEnd's Kernel) get this Package and I want
> to answer. But obviously the answer is not recognized. Am I talking the
> wrong language to the Kernel?
> I write directly (i.e. without a package "around" the command)
> LinkWrite[theLink,"retry"]
> This works for example for
> LinkInterrupt[theLink];
> LinkWrite[theLink,"continue"];
> Shouldn't it also work for this problem?
> Thanks,
> Manuel
> On Wed, 2005-11-23 at 01:12 -0500, Yasvir A. Tesiram wrote:
>> Hi,
>> If MathLM crashes, the only thing you can do is to restart MathLM. Once
>> the message "...could not contact the MathLM server..." pops up, you have
>> no way to use Mathematica. You have to restart MathLM first before you can
>> launch/relaunch Mathematica.
>> Any connections you have to remote kernels from that first instance should
>> also pop a message on your display, and after MathLM is relaunched just
>> type "r" (for the command line interface) or push the button that says
>> "retry" (for the Notebook interface).
>> Maybe you need to talk to your system administrator and find out why
>> MathLM crashes and maybe he/she can do something to restart it
>> automatically after automated reboots, or power failues etc.
>> Regards
>> Yas
>> On Tue, 22 Nov 2005, Manuel Schmidt wrote:
>>> Hi,
>>> Thank you for the answer. Here are some points:
>>> MathLM is running on a server and I run 20 kernels on different
>>> computers. I cannot log into the server, because I am just a user.
>>> I know, that mathlm crashes from time to time and I think, there is
>>> nothing, I can change about that. It is not a connection issue, because
>>> those computers are linked directly by GB-Ethernet.
>>> My problem is a different one:
>>> If the connection to MathLM is lost, then normally your kernel asks you
>>> if it should retry the server by a dialog box.
>>> I link all those kernels to a local kernel via mathlink and give
>>> commands via LinkWrite.
>>> My problem is, that if the connection is lost, I get via LinkRead this
>>> MenuPacket[2, Lost connection to server> ]
>>> and I would think, that I have to do a LinkWrite[theLink,"retry"] to
>>> tell the kernel, he should reconnect to MathLM.
>>> However, this does not work. Everytime I send a "retry" or "r" to the
>>> Kernel, I get back a new
>>> MenuPacket[2, Lost connection to server> ]
>>> .
>>> What should I send the kernel?
>>> Thanks,
>>> Manuel
>>> On Sun, 2005-11-20 at 16:58 -0600, Yasvir A. Tesiram wrote:
>>>> Hi,
>>>> It is hard to know. Is MathLM running on the same computer that you are
>>>> running MathKernel? If so, check to see that you haven't inadvertently
>>>> set yourself on a timed lease.
>>>> If MathLM is on a remote computer, you should try and log on to the
>>>> computer that serves the license and check the log file (assuming you
>>>> write one) to see what happened. Then restart it, assuming you are the
>>>> system administrator and assuming that MathLM had crashed. When this
>>>> happens, is the server still running? Can you log on to it?
>>>> There is also the issue of connection quality. So, if the connection is of
>>>> poor quality, you will lose signal from time to time.
>>>> What version of MathLM is it?
>>>> What OS is it running on?
>>>> Are you the only user?
>>>> Regards
>>>> Yas
>>>> On Sat, 19 Nov 2005, Manuel Schmidt wrote:
>>>>> Hi,
>>>>> I run a MathKernel over MathLink with a License from a MathLM. Sometimes
>>>>> this licenseserver crashes and I get through the link a
>>>>> MenuPacket[2, Lost connection to server> ]
>>>>> now with LinkWrite[theLink,"retry"] i want to tell the kernel to
>>>>> reconnect to MathLM, but I get again the same message as above.
>>>>> What am I doing wrong?
>>>>> Thanks,
>>>>> Manuel

  • Prev by Date: Re: Re: Lost connection to license server
  • Next by Date: Re: Types in Mathematica
  • Previous by thread: Re: Re: Lost connection to license server
  • Next by thread: Same scaling for plots/ charts