Embedded calls to `find` in cleanUsersData.php do not scale up

[jmfernandez]

I have found in the logs emitted by cleanUsersData.php from other openVRE deployments messages like next:

Looking for temporary data (/gpfs/vre/userdata//*/*/.tmp/)
-- find /gpfs/vre/userdata//*/*/.tmp/  -maxdepth 1  -mtime +7 -type f  -exec rm -fv {} \;
sh: 1: find: Argument list too long
-- find /gpfs/vre/userdata//*/*/.tmp/  -maxdepth 1  -mtime +7 -type d -name '?*_[0-9]*'  -exec rm -Rfv {} \;
sh: 1: find: Argument list too long
-- find /gpfs/vre/userdata//*/*/.tmp/  -maxdepth 1  -size +2G -mtime -7 -type f -name '*tar.gz'  -exec rm -fv {} \;
sh: 1: find: Argument list too long

And they are happening when the user data directory from that openVRE deployment contains too many entries (more than the number of parameters which can passed to a program from command line). These are the blamed lines:

openVRE-core-dev/front_end/openVRE/scripts/maintainance/cleanUsersData.php

Lines 341 to 363 in d8d98c8

	//$cmd="find ".$GLOBALS['dataDir']."/*/*/".$GLOBALS['tmpUser_dir']." -maxdepth 1 -mtime +".$caduca_strict." -type f -exec rm -fv {} \\;";
	$cmd = "find ".$GLOBALS['dataDir']."/*/*/".$GLOBALS['tmpUser_dir']." -maxdepth 1 -mtime +".$caduca_strict." -type f ";
	if ($dry_run === false) {
	$cmd.= " -exec rm -fv {} \\;";
	}
	print "-- $cmd\n";
	system($cmd);
	// find and delete directories in .tmp named like '?*_[0-9]*' and older than 'caduca_strict'
	//$cmd="find ".$GLOBALS['dataDir']."/*/*/".$GLOBALS['tmpUser_dir']." -maxdepth 1 -mtime +".$caduca_strict." -type d -name '?*_[0-9]*' -exec rm -Rfv {} \\;";
	$cmd = "find ".$GLOBALS['dataDir']."/*/*/".$GLOBALS['tmpUser_dir']." -maxdepth 1 -mtime +".$caduca_strict." -type d -name '?*_[0-9]*' ";
	if ($dry_run === false) {
	$cmd.= " -exec rm -Rfv {} \\;";
	}
	print "-- $cmd\n";
	system($cmd);
	// find and delete TAR files in .tmp bigger than 2GB // TODO: no need of it when issue with 'massive TARs' is solved
	//$cmd="find ".$GLOBALS['dataDir']."/*/*/".$GLOBALS['tmpUser_dir']." -maxdepth 1 -size +2G -mtime -".$caduca_strict." -type f -name '*tar.gz' -exec rm -fv {} \\;";
	$cmd = "find ".$GLOBALS['dataDir']."/*/*/".$GLOBALS['tmpUser_dir']." -maxdepth 1 -size +2G -mtime -".$caduca_strict." -type f -name '*tar.gz' ";
	if ($dry_run === false) {
	$cmd.= " -exec rm -fv {} \\;";
	}

This issue can be fixed using something similar to next command line, where find command delegates on xargs calling nested find commands, so no wildcard expansion happens:

    if(substr($GLOBALS['tmpUser_dir'], -1) == "/") {
        $tmpuser_dir = substr($GLOBALS['tmpUser_dir'], 0, -1);
    } else {
        $tmpuser_dir = $GLOBALS['tmpUser_dir'];
    }

    $cmd = "find ".$GLOBALS['dataDir']." -type d -name ".$tmpuser_dir." -print0 | xargs -0 -I{.} find {.} -maxdepth 1  -mtime +".$caduca_strict." -type f ";
    if ($dry_run === false) {
        $cmd.= " -exec rm -fv {} \\;";
    }
    print "-- $cmd\n";
    system($cmd);

    $cmd = "find ".$GLOBALS['dataDir']." -type d -name ".$tmpuser_dir." -print0 | xargs -0 -I{.} find {.} -maxdepth 1  -mtime +".$caduca_strict." -type d -name '?*_[0-9]*' ";
    if ($dry_run === false) {
        $cmd.= " -exec rm -Rfv {} \\;";
    }
    print "-- $cmd\n";
    system($cmd);

    $cmd = "find ".$GLOBALS['dataDir']." -type d -name ".$tmpuser_dir." -print0 | xargs -0 -I{.} find {.} -maxdepth 1  -size +2G -mtime -".$caduca_strict." -type f -name '*tar.gz' ";
    if ($dry_run === false) {
        $cmd.= " -exec rm -fv {} \\;";
    }
    print "-- $cmd\n";
    system($cmd);