Add plab/stale_age sitevar to control what we consider stale data.

If set to zero, we never consider data stale, we use it anyway.
Nodes which don't report any data are still considered unusable.

Also, consider nodes that don't report disk usage as 100% free so that
we will consider them.

tbsetup/plab/plabmetrics.in

@@ -15,8 +15,9 @@ my $STALESLOP = 1 * 60;
 
 #
 # Age (in seconds) at which we consider metric data stale
+# Configured via site variable.
 #
-my $STALEAGE  = 30 * 60;
+my $STALEAGE;
 
 sub usage()
 {
@@ -138,6 +139,14 @@ if (TBSiteVarExists("plab/min_disk")) {
     $MINDISK = 10;
 }
 
+if (TBSiteVarExists("plab/stale_age")) {
+    $STALEAGE = TBGetSiteVar("plab/stale_age");
+    $STALEAGE = 0 if $STALEAGE < 0;
+} else {
+    $STALEAGE = 60;
+}
+$STALEAGE *= 60;
+
 print "\n=== plabmetrics ".
     "(metric=$LOADMETRIC, maxload=$MAXLOAD, mindisk=$MINDISK) ".
     "running at " . `date`
@@ -243,8 +252,15 @@ sub CheckMetrics()
 	$localdebug = 0;
     } elsif ($metricsage < 0) {
 	if (-$metricsage > $STALESLOP) {
+	    print "WARNING: $host: metric data in the future, ignoring\n"
+		if $localdebug;
 	    $metrics{$LOADMETRIC} = 999;
 	}
+    } elsif ($STALEAGE == 0) {
+	if ($metricsage > 4 * 60 * 60) {
+	    print "WARNING: $host: metric data older than 4 hours, using anyway\n"
+		if $localdebug;
+	}
     } elsif ($metricsage > $STALEAGE) {
 	print "WARNING: $host: stale metric data, ignoring\n"
 	    if $localdebug;
@@ -259,18 +275,15 @@ sub CheckMetrics()
 	    if $localdebug;
 	$metrics{$LOADMETRIC} = 999;
     }
-    if (!defined($metrics{disk_free})) {
-	print "WARNING: $host: no disk_free metric\n"
-	    if $localdebug;
-	$metrics{disk_free} = 0;
-    }
-    if (!defined($metrics{disk_total})) {
-	print "WARNING: $host: no disk_total metric\n"
+    if (!defined($metrics{disk_free}) || !defined($metrics{disk_total})) {
+	print "WARNING: $host: no disk_free/disk_total metrics, assuming enough\n"
 	    if $localdebug;
-	$metrics{disk_total} = $metrics{disk_free};
+	$metrics{disk_free} = $metrics{disk_total} = 1;
     }
     $metrics{disk_total} = 1
 	if $metrics{disk_total} == 0;
+    $metrics{disk_free} = $metrics{disk_total}
+	if $metrics{disk_free} > $metrics{disk_total};
 }
 
 #

tbsetup/plab/plabstats.in

@@ -387,18 +387,15 @@ sub CheckMetrics()
 	$metrics{bytes_out} = 0;
     }
 
-    if (!defined($metrics{disk_free})) {
-	print "* * * WARNING $host: no disk_free metric\n"
+    if (!defined($metrics{disk_free}) || !defined($metrics{disk_total})) {
+	print "* * * WARNING: $host: no disk_free/disk_total metrics\n"
 	    if $debug;
-	$metrics{disk_free} = 0;
-    }
-    if (!defined($metrics{disk_total})) {
-	print "* * * WARNING $host: no disk_total metric\n"
-	    if $debug;
-	$metrics{disk_total} = $metrics{disk_free};
+	$metrics{disk_free} = $metrics{disk_total} = 1;
     }
     $metrics{disk_total} = 1
 	if $metrics{disk_total} == 0;
+    $metrics{disk_free} = $metrics{disk_total}
+	if $metrics{disk_free} > $metrics{disk_total};
 }
 
 #