I have production script that copies and renames files from one mounted network share to another (both CIFS), and sometimes the copy function hangs indefinitely on read I/O dropping the process into uninterruptible sleep where it can only be killed with SIGKILL. (Thankfully, TASK_KILLABLE, apparently.)
Since operation is blocked at the copy call I can't gracefully handle the I/O failure, nor, more importantly, log it.
$res = copy("/path/to/mount/file.pdf", "/path/to/productionqueue/newfile.pdf");
//This doesn't run because of the process state.
if($res) {
//Report success to the log.
} else {
//Report failure to the log.
}
It's an environment, file-specific issue, and manual intervention can fix it, but I need to log the failure so an administrator can be notified about the condition. (Aside: I think it's a lock race condition. I can fix the problem by unmounting and remounting the source share, but reproducing it is a bit of an issue.)
Ideally, I'd like the copy call to timeout after 10 seconds, so I can log the error, but there doesn't seem to be a way in PHP to do that.
What I'm considering is delegating copy or read operations to a subprocess monitored by the timeout command and responding based on the exit codes, but is there a simpler way in a PHP script to handle this sort of scenario?
Everything I've found in my search suggests there's only runtime configuration available for socket/network timeouts, not filesystem calls.
