* Load balancing. `build-remote.pl' will only execute up to a

  per-machine maximum number of parallel jobs on a remote machine.

4 files changed, 34 insertions, 11 deletions

diff --git a/distributed/build-remote.pl b/distributed/build-remote.pl
index f1a9b10f4..12cb272b9 100755
--- a/distributed/build-remote.pl
+++ b/distributed/build-remote.pl
@@ -7,9 +7,16 @@ my $localSystem = shift @ARGV;
 my $neededSystem = shift @ARGV;
 my $storeExpr = shift @ARGV;
 
+sub sendReply {
+    my $reply = shift;
+    open OUT, ">&3" or die;
+    print OUT "$reply\n";
+    close OUT;
+}
+
 # Decline if the local system can do the build.
 if ($amWilling && ($localSystem eq $neededSystem)) {
-    print "decline\n";
+    sendReply "decline";
     exit 0;
 }
 
@@ -45,11 +52,8 @@ while (<LOAD>) {
 }
 close LOAD;
 
-sub sendReply {
-    my $reply = shift;
-    open OUT, ">&3" or die;
-    print OUT "$reply\n";
-    close OUT;
+foreach my $cur (keys %machines) {
+    $curJobs{$cur} = 0 unless defined $curJobs{$cur};
 }
 
 # Find a suitable system.
@@ -58,8 +62,7 @@ my $machine;
 foreach my $cur (keys %machines) {
     if ($neededSystem eq $systemTypes{$cur}) {
         $rightType = 1;
-        if (!defined $curJobs{$cur} or
-            ($curJobs{$cur} < $maxJobs{$cur}))
+        if ($curJobs{$cur} < $maxJobs{$cur})
         {
             $machine = $cur;
             last;
@@ -86,6 +89,19 @@ close IN;
 
 print "BUILDING REMOTE: $storeExpr on $machine\n";
 
+$curJobs{$machine} = $curJobs{$machine} + 1;
+
+sub writeLoad {
+    open LOAD, "> /home/eelco/nix/distributed/current-load" or die;
+    foreach my $cur (keys %machines) {
+        print LOAD "$cur $curJobs{$cur}\n";
+    }
+    close LOAD;
+}
+
+writeLoad
+
+
 my $ssh = "ssh -i $sshKeys{$machine} -x";
 
 my $inputs = `cat inputs`;
@@ -117,3 +133,7 @@ foreach my $output (split '\n', $outputs) {
     system "rsync -a -e '$ssh' $machine:$output /nix/store";
     die "cannot rsync outputs from $machine" if ($? != 0);
 }
+
+$curJobs{$machine} = $curJobs{$machine} - 1;
+
+writeLoad
diff --git a/distributed/remote-systems.conf b/distributed/remote-systems.conf
index 4e6d2b962..3c9abc719 100644
--- a/distributed/remote-systems.conf
+++ b/distributed/remote-systems.conf
@@ -1,2 +1,2 @@
-scratchy.labs.cs.uu.nl   i686-linux     /home/eelco/.ssh/id_scratchy_auto       2
+scratchy.labs.cs.uu.nl   i686-linux     /home/eelco/.ssh/id_scratchy_auto       1
 mckroket.labs.cs.uu.nl   powerpc-mac    /home/eelco/.ssh/id_scratchy_mckroket   1
diff --git a/distributed/test.sh b/distributed/test.sh
index 38116eba5..78f83095e 100755
--- a/distributed/test.sh
+++ b/distributed/test.sh
@@ -2,8 +2,11 @@
 
 set -e
 
+rm -f current-load
+touch current-load
+
 storeExpr=$(nix-instantiate ~/nixpkgs/pkgs/system/test.nix)
 
 export NIX_BUILD_HOOK="build-remote.pl"
 
-../src/nix-store/nix-store -qnvvvv -j0 $storeExpr
+../src/nix-store/nix-store -qnvvvv -j1 $storeExpr
diff --git a/src/libstore/normalise.cc b/src/libstore/normalise.cc
index 3ae3fb789..0673814ba 100644
--- a/src/libstore/normalise.cc
+++ b/src/libstore/normalise.cc
@@ -332,7 +332,7 @@ void Normaliser::run()
 
     while (!goals.empty()) {
 
-        printMsg(lvlVomit, "main loop");
+        debug("main loop - starting jobs");
         
         /* Start building as many buildable goals as possible. */
         bool madeProgress = false;