couldn't load file "/viper/u1/system/soft/RHEL_9/packages/x86_64/Modules/5.4.0/lib/libtclenvmodules.so"

[xushanthu-2014]

Describe the bug

When starting a shell/login session on the MPCDF Viper HPC cluster from Visual Studio Code, Environment Modules fails during initialization because it cannot load libtclenvmodules.so.

The error appears immediately after login / when opening a VS Code remote terminal, or using ssh in Mac terminal:

ERROR: couldn't load file "/viper/u1/system/soft/RHEL_9/packages/x86_64/Modules/5.4.0/lib/libtclenvmodules.so": /viper/u1/system/soft/RHEL_9/packages/x86_64/Modules/5.4.0/lib/libtclenvmodules.so: cannot read file data: Input/output error
      while executing
  "load [file normalize [getConf tcl_ext_lib]] Envmodules"
  Please report this issue at https://github.com/cea-hpc/modules/issues

The file metadata is visible, but reading the file contents fails:

$ ls -l /viper/u1/system/soft/RHEL_9/packages/x86_64/Modules/5.4.0/lib/libtclenvmodules.so
-rwxr-xr-x 1 root hpcsw 25608 Jun  2 10:49 /viper/u1/system/soft/RHEL_9/packages/x86_64/Modules/5.4.0/lib/libtclenvmodules.so

$ md5sum /viper/u1/system/soft/RHEL_9/packages/x86_64/Modules/5.4.0/lib/libtclenvmodules.so
md5sum: /viper/u1/system/soft/RHEL_9/packages/x86_64/Modules/5.4.0/lib/libtclenvmodules.so: Input/output error

To Reproduce

Steps to reproduce the behavior:

$ ssh viper01
# After login, the error is printed during shell/module initialization.

or by opening a VS Code Remote-SSH terminal on the same host.

Running md5sum on the shared library also reproduces the underlying read failure:

$ md5sum /viper/u1/system/soft/RHEL_9/packages/x86_64/Modules/5.4.0/lib/libtclenvmodules.so
md5sum: /viper/u1/system/soft/RHEL_9/packages/x86_64/Modules/5.4.0/lib/libtclenvmodules.so: Input/output error

Location and content of any modulerc or modulefile involved:

Not sure. The failure happens before I can successfully use module, while Modules is loading its Tcl extension library.

Expected behavior

Modules should initialize normally without printing an error. Commands such as module avail, module load, and terminal startup should work normally.

Error and debugging information

The initialization error is:

ERROR: couldn't load file "/viper/u1/system/soft/RHEL_9/packages/x86_64/Modules/5.4.0/lib/libtclenvmodules.so": /viper/u1/system/soft/RHEL_9/packages/x86_64/Modules/5.4.0/lib/libtclenvmodules.so: cannot read file data: Input/output error
      while executing
  "load [file normalize [getConf tcl_ext_lib]] Envmodules"
  Please report this issue at https://github.com/cea-hpc/modules/issues

The shared library exists according to ls, but file content reads fail:

$ ls -l /viper/u1/system/soft/RHEL_9/packages/x86_64/Modules/5.4.0/lib/libtclenvmodules.so
-rwxr-xr-x 1 root hpcsw 25608 Jun  2 10:49 /viper/u1/system/soft/RHEL_9/packages/x86_64/Modules/5.4.0/lib/libtclenvmodules.so

$ md5sum /viper/u1/system/soft/RHEL_9/packages/x86_64/Modules/5.4.0/lib/libtclenvmodules.so
md5sum: /viper/u1/system/soft/RHEL_9/packages/x86_64/Modules/5.4.0/lib/libtclenvmodules.so: Input/output error

Modules version and configuration

I could not reliably obtain this because Modules fails while loading libtclenvmodules.so.

If useful, I can provide the output of these commands once the initialization works again:

$ module --version
$ module config --dump-state

Additional context

This happened on the MPCDF Viper HPC cluster. I connected using ssh viper01, and the actual login node shown in the prompt was viper02.

Because md5sum on libtclenvmodules.so returns Input/output error, this may be a filesystem/storage issue rather than a Modules logic bug. I am reporting it here because the error message explicitly says to report the issue at this repository.

[xdelaruelle]

Many thanks for your report @xushanthu-2014.

Please also communicate this issue to the MPCDF support team as it appears to be an I/O issue on their infrastructure.

On our side, we will fix the error message, not to indicate that such error should be reported to Modules developers.