From a3f95cc0e2fb3c359fa989f58691ad44b39173eb Mon Sep 17 00:00:00 2001
From: Martin Mares <mj@ucw.cz>
Date: Mon, 31 Oct 2011 14:37:58 +0100
Subject: [PATCH] Implemented preparatory commands

---
 NOTES |  5 ++++
 TODO  |  1 -
 brun  | 87 ++++++++++++++++++++++++++++++++++++++++++++++-------------
 3 files changed, 74 insertions(+), 19 deletions(-)

diff --git a/NOTES b/NOTES
index 561abd6..9f3dc17 100644
--- a/NOTES
+++ b/NOTES
@@ -32,6 +32,9 @@ Known header fields:
 
 ID: <job-id>			Identifier of the job, unique in the scope of a queue
 Subject: <subject>		Subject to be displayed to the user
+Prep: <command>			Run <command> in a shell before the job body is executed;
+				$HOST contains the name of the host. This is useful for
+				example if you want to transfer data to the host by rsync.
 
 ### Status files ###
 
@@ -52,11 +55,13 @@ OK		Job finished successfully (this is usually not seen in the queue, since
 		finished jobs are immediately deleted or moved to the history)
 FAILED		Job failed to execute (i.e., it returned a non-zero exit code)
 INTERR		Internal error of BEX (e.g., failed to read job prolog file)
+PREPFAIL	Preparatory commands failed (i.e., those present in Prep header field)
 
 These are present only in log files and messages sent over status FIFO:
 
 RUN		Job is running
 SEND		Sending job to the host
+PREP		Running preparatory commands (i.e., those present in Prep header field)
 
 Status codes not tied to a specific job (sent over status FIFO):
 
diff --git a/TODO b/TODO
index b2a374d..b7badfd 100644
--- a/TODO
+++ b/TODO
@@ -1,7 +1,6 @@
 - bprun: option for setting max # of running jobs
 - bprun --job
 - bprun --curses
-- rsync, rsync-only
 - Terminology: machine vs. host
 - ssh options
 - Detector of orphans (unused queue dirs, jobs on non-existent machines, non-queued jobs)
diff --git a/brun b/brun
index 324f23f..3e46d47 100755
--- a/brun
+++ b/brun
@@ -42,19 +42,64 @@ sub update_status($$$$;$) {
 	}
 }
 
+my %pings;
+
 sub ping_machine($) {
 	my ($mach) = @_;
-	return 1 unless $BEX::Config::ping_hosts;
-	update_status($mach, '-', 'PING', undef);
-	my $host = BEX::Config::host_name($mach);
-	`ping -c1 -n $host >/dev/null 2>/dev/null`;
-	return !$?;
+	if (!defined $pings{$mach}) {
+		if ($BEX::Config::ping_hosts) {
+			update_status($mach, '-', 'PING', undef);
+			my $host = BEX::Config::host_name($mach);
+			`ping -c1 -n $host >/dev/null 2>/dev/null`;
+			$pings{$mach} = !$?;
+		} else {
+			$pings{$mach} = 1;
+		}
+	}
+	if ($pings{$mach}) {
+		return ('OK', undef);
+	} else {
+		return ('NOPING', 'Does not ping');
+	}
 }
 
-sub run_job($$$) {
+sub exit_status($) {
+	my ($s) = @_;
+	if ($s >> 8) {
+		return "with exit code " . ($s >> 8);
+	} else {
+		return "on fatal signal " . ($s & 127);
+	}
+}
+
+sub run_job_prep($$$) {
+	my ($job, $queue, $mach) = @_;
+	my $prep = $job->attr('Prep');
+	defined($prep) && $prep !~ /^\s*$/ or return 'OK';
+
+	my $jid = $job->id;
+	update_status($mach, $jid, 'PREP', $queue);
+	my $lf = $queue->log_file($mach, $jid);
+	$ENV{'HOST'} = BEX::Config::host_name($mach);
+	system 'bash', '-o', 'pipefail', '-c', "( $prep ) 2>&1 | tee -a $lf";
+	delete $ENV{'HOST'};
+	if ($?) {
+		return ('PREPFAIL', 'Preparatory command failed ' . exit_status($?));
+	} else {
+		return 'OK';
+	}
+}
+
+sub run_job_body($$$) {
 	my ($job, $queue, $mach) = @_;
-	my $jid = $job->{'ID'};
+
+	if ($job->attr('body') =~ /^\s*$/s) {
+		# Shortcut if the body is empty
+		return 'OK'
+	}
+
 	my $host = BEX::Config::host_name($mach);
+	my $jid = $job->id;
 
 	my $tmp = $queue->temp_file($mach, $jid);
 	open T, '>', $tmp or die;
@@ -66,7 +111,7 @@ sub run_job($$$) {
 		print T "#!/bin/sh\n";
 	}
 	print T "# BEX job ", $jid, "\n";
-	print T $job->{'body'};
+	print T $job->attr('body');
 	if (defined $BEX::Config::job_epilog) {
 		open E, $BEX::Config::job_epilog or return ('INTERR', "Cannot open epilog: $!");
 		while (<E>) { print T; }
@@ -84,12 +129,25 @@ sub run_job($$$) {
 	my $lf = $queue->log_file($mach, $jid);
 	system 'bash', '-o', 'pipefail', '-c', "ssh -t $host '$rtmp ; e=\$? ; rm -f $rtmp ; exit \$e' 2>&1 | tee -a $lf";
 	if ($?) {
-		return ('FAILED', 'Job failed');
+		return ('FAILED', 'Job failed ' . exit_status($?));
 	} else {
-		return ('OK', undef);
+		return 'OK';
 	}
 }
 
+sub run_job($$$) {
+	my ($job, $queue, $mach) = @_;
+	my ($stat, $msg);
+
+	($stat, $msg) = ping_machine($mach);
+	$stat eq 'OK' or return ($stat, $msg);
+
+	($stat, $msg) = run_job_prep($job, $queue, $mach);
+	$stat eq 'OK' or return ($stat, $msg);
+
+	return run_job_body($job, $queue, $mach);
+}
+
 my @machines = BEX::Config::parse_machine_list(@ARGV ? @ARGV : '*');
 my $queue = BEX::Queue->new($queue_name);
 
@@ -104,7 +162,6 @@ for my $mach (@machines) {
 		next;
 	}
 	update_status($mach, '-', 'INIT', undef);
-	my $ping;
 	while (my $jid = shift @q) {
 		if (defined $given_job) {
 			$jid eq $given_job or next;
@@ -119,13 +176,7 @@ for my $mach (@machines) {
 			'Time' => time,
 		};
 		print "### Running $jid (", $job->attr('Subject'), ") on $mach ###\n";
-		$ping //= ping_machine($mach);
-		my ($s, $msg);
-		if (!$ping) {
-			($s, $msg) = ('NOPING', 'Does not ping');
-		} else {
-			($s, $msg) = run_job($job, $queue, $mach);
-		}
+		my ($s, $msg) = run_job($job, $queue, $mach);
 
 		$stat->{'Status'} = $s;
 		$stat->{'Message'} = $msg;
-- 
2.39.2