Forum Discussion

FICO's avatar
FICO
Community Manager
3 years ago

Error: Unexpected failure accessing global file / no available space on data medium

Hi everyone,

I try to solve a huge capacitated facility location problem. After the run time of 2 days and still no optimal solution the following error occurred: ?45 Error: Unexpected failure accessing global file - Error accessing archive file: There is not enough space available on the data medium.

I am working with Xpress IVE Version 1.25.06 64 bit, Mosel Version 6.0.1.

Can anyone please explain to me why such huge amount of data is written to my computers data medium while running the model (over 30 GB)? And especially how to fix this error?

Thank you!

Yvonne

4 Replies

  • FICO's avatar
    FICO
    Community Manager

    Hi Yvonne,

    when the search tree gets too large, Xpress optimizer starts swapping out data to disk to avoid out of memory situations.

    One way to solve your problem would be to forbid this swapping and also set a maximum memory limit. If that limit is hit then the optimizer exits gracefully, just as if a time limit, node limit, ... is hit.

    The MAXMEMORYHARD control allows limiting the total amount of memory Xpress will allocate.

    The TREEMEMORYLIMIT control tells the optimizer when to start swapping out to disk. If you set this to a very large value (like 2147483647, the maximum signed integer) then no swapping to disk should happen.

  • FICO's avatar
    FICO
    Community Manager

    Hi Daniel,

    thank you so much for your quick and helpful answer! :)

    Do you also know how to set to which disk the data swapping should happen? 

  • FICO's avatar
    FICO
    Community Manager

    The name of the swap file is derived from the model name. The model name can be set using XPRSsetprobname for the optimizer or XPRS_probname in Mosel.

    If you set this to "/tmp/model" then your tree file should be created in the "/tmp" directory.

    Yes, this is rather obscure.

  • FICO's avatar
    FICO
    Community Manager

    Great, thank you so much!