Context

I get inconsistent delays (from a fraction of a second to 10 seconds or more) in launching

'echo "Print[1+1]; Quit[]" | math'  

on different nodes of a cluster (or at different times on the same node), which seems to be driven by the fact that these nodes do not have access to the internet. It seems that the kernels are sometimes trying to access wolfram servers, hence the delay. The delay disappears if I ask the system manager to temporarily open such access.

Question

How to tell Mathematica not to try and talk to wolfram?

Note

This question does not necessarily apply to only subKernels. It can be the main kernel as well.

Context

I get inconsistent delays (from a fraction of a second to 10 seconds or more) in launching

'echo "Print[1+1]; Quit[]" | math'  

on different nodes of a cluster (or at different times on the same node), which seems to be driven by the fact that these nodes do not have access to the internet. It seems that the kernels are sometimes trying to access wolfram servers, hence the delay. The delay disappears if I ask the system manager to temporarily open such access.

Question

How to tell Mathematica not to try and talk to wolfram?

As an artefact, it would also delay considerably the frontEnd, should one attempt to run it on such a compute node.

Notes

1. This question does not necessarily apply to only subKernels. It can be the main kernel as well.
2. It could also be a licensing issue, though it seems unlikely because eventually it works.
3. This problem occurs with mathematica 8,9, 10.3, 11.3 and 12.

Context

I get inconsistent delays (from a fraction of a second to 10 seconds or more) in launching

'echo "Print[1+1]; Quit[]" | math'  

on different nodes of a cluster (or at different times on the same node), which seems to be driven by the fact that these nodes do not have access to the internet. It seems that the kernels are sometimes trying to access wolfram servers, hence the delay. The delay disappears if I ask the system manager to temporarily open such access.

Question

How to tell Mathematica not to try and talk to wolfram?

Context

I get inconsistent delays (from a fraction of a second to 10 seconds or more) in launching

'echo "Print[1+1]; Quit[]" | math'  

on different nodes of a cluster (or at different times on the same node), which seems to be driven by the fact that these nodes do not have access to the internet. It seems that the kernels are sometimes trying to access wolfram servers, hence the delay. The delay disappears if I ask the system manager to temporarily open such access.

Question

How to tell Mathematica not to try and talk to wolfram?

Note

This question does not necessarily apply to only subKernels. It can be the main kernel as well.

Context

I get inconsistent delays (from a fraction of a second to 10 seconds or more) in launching

'echo "Print[1+1]; Quit[]" | math'  

on different nodes of a cluster (or at different times on the same node), which seems to be driven by the fact that these nodes do not have access to the internet. It seems that the kernels are sometimes trying to access wolfram, hence the delay. The delay disappears if I ask the system manager to temporarily open such access.

Question

How to tell Mathematica not to try and talk to wolfram?

Context

I get inconsistent delays (from a fraction of a second to 10 seconds or more) in launching

'echo "Print[1+1]; Quit[]" | math'  

on different nodes of a cluster (or at different times on the same node), which seems to be driven by the fact that these nodes do not have access to the internet. It seems that the kernels are sometimes trying to access wolfram servers, hence the delay. The delay disappears if I ask the system manager to temporarily open such access.

Question

How to tell Mathematica not to try and talk to wolfram?