Originally posted by: RayZ

Is there a way to turn off CPLEX's log file from the Matlab interface? I have not found a way to avoid CPLEX attempting to create a clone1.log file every time I run a barrier method solve.

Which version of CPLEX are you using?  Are you using the Toolbox or Class API?  Did you set the clonelog parameter?

Originally posted by: RayZ

CPLEX 12.6 and I'm using the Toolbox. I did not set the clonelog parameter. My understanding was that only affects whether the log file is cloned for multiple threads, but cannot be used to turn it off completely.

I'm not very familar with the matlab API's, but if you set the clonelog parameter to -1 you should not get any clone*.log files.  Are you saying that a cplex.log file is also being created on your file system?  I've run a few matlab examples and I'm not seeing this behavior....

Originally posted by: RayZ

Even if I set the clonelog parameter to -1, I am still getting a clone1.log file (no clpex.log file). The attached file, containing the code below, reproduces the problem consistently for me. Maybe there is something wrong with the way I'm setting the parameter?

I was able to reproduce this with 12.6.0.0, but it appears to be fixed with 12.6.0.1.  See info on the fix pack here.

Originally posted by: RayZ

Thanks. I have confirmed that 12.6.0.1 fixes the issue for me as well.

Originally posted by: ThomasPfau

Is it possible, that this fix was patched out when 12.6.1 was released?

Because I do get those logs occasionally) regardless on whether I set the appropriate parameter...
I don't get them on small problems so its not so easy to get a good example, but they do appear despite the parameter being -1.

Can someone confirm this?

EDIT: Interstingly, the parameter talks about specification of a log file, and I can not find any option to do so. Similarily,  the description of the parallel optimizer, talkes about it using a log file if there is an active log file. But I wouldn't even know how to figure ou whether there is one or how to deactivate it.

EDIT2: I just tested it with 12.6.0.1 as well, and I do get a logfile with that version as well. So this seems not to have been fixed. These log files are particularly frustrating when I to compute things on a cluster, as there is no way to at least indicate which folders to use for the log files. Thus, this is bound to create signifiicant amounts of network usage when running any program form a path that is not on a local node but on a nfs..
 

I could not reproduce this issue here. For me disabling the clone logs works without a problem. Could you send/post a minimal example code to reproduce this issue?

Concerning your first edit: The callable library (on which the matlab connector is based) uses log files, that is why the parameter description talks about log files. The matlab connector however does not use the logfile feature of the callable library. Thus, in matlab, the parameter has in fact only two valid values.