I use tail to monitor the progress of jobs that I know will write their progress to disk. Almost always, I know which file they will create before they start running (the jobs are dispatched by a scheduler on a supercomputer)
Is there a way to tail these files before they are created? I would like to do so while avoiding race conditions and/or making assumptions about how or when
the jobs write to disk.
Use the -F flag to tail (assuming you have tail from GNU coreutils):
tail -F file-that-does-not-exist
From man tail:
-F same as --follow=name --retry
--retry
keep trying to open a file even when it is or becomes inaccessi‐
ble; useful when following by name, i.e., with --follow=name
-f, --follow[={name|descriptor}]
output appended data as the file grows; -f, --follow, and --fol‐
low=descriptor are equivalent
If your tail implementation does not have the --retry option, you could probably just cheat and create the file yourself:
$ touch file && tail-f file
Depending on your requirements with respect to ownership and permissions that might be sufficient.
