display all backups which we don't know state of in web interface
[BackupPC.git] / bin / BackupPC
index 585eec1..c6fc7d2 100755 (executable)
@@ -1,4 +1,4 @@
-#!/bin/perl
+#!/usr/bin/perl
 #============================================================= -*-perl-*-
 #
 # BackupPC: Main program for PC backups.
@@ -6,7 +6,7 @@
 # DESCRIPTION
 #
 #   BackupPC reads the configuration and status information from
-#   $TopDir/conf.  It then runs and manages all the backup activity.
+#   $ConfDir/conf.  It then runs and manages all the backup activity.
 #
 #   As specified by $Conf{WakeupSchedule}, BackupPC wakes up periodically
 #   to queue backups on all the PCs.  This is a four step process:
@@ -29,7 +29,7 @@
 #   Craig Barratt  <cbarratt@users.sourceforge.net>
 #
 # COPYRIGHT
-#   Copyright (C) 2001-2003  Craig Barratt
+#   Copyright (C) 2001-2009  Craig Barratt
 #
 #   This program is free software; you can redistribute it and/or modify
 #   it under the terms of the GNU General Public License as published by
@@ -47,7 +47,7 @@
 #
 #========================================================================
 #
-# Version 2.1.0beta1, released 9 Apr 2004.
+# Version 3.2.0, released 31 Jul 2010.
 #
 # See http://backuppc.sourceforge.net.
 #
@@ -59,6 +59,7 @@ use vars qw(%Status %Info $Hosts);
 use lib "/usr/local/BackupPC/lib";
 use BackupPC::Lib;
 use BackupPC::FileZIO;
+use Encode qw/decode_utf8/;
 
 use File::Path;
 use Data::Dumper;
@@ -66,6 +67,7 @@ use Getopt::Std;
 use Socket;
 use Carp;
 use Digest::MD5;
+use POSIX qw(setsid);
 
 ###########################################################################
 # Handle command line options
@@ -86,6 +88,7 @@ if ( !getopts("d", \%opts) || @ARGV != 0 ) {
 die("BackupPC::Lib->new failed\n") if ( !(my $bpc = BackupPC::Lib->new) );
 my $TopDir = $bpc->TopDir();
 my $BinDir = $bpc->BinDir();
+my $LogDir = $bpc->LogDir();
 my %Conf   = $bpc->Conf();
 
 #
@@ -110,10 +113,14 @@ if ( $Conf{BackupPCUserVerify}
 #
 # Read old status
 #
-if ( -f "$TopDir/log/status.pl" && !(my $ret = do "$TopDir/log/status.pl") ) {
-   die "couldn't parse $TopDir/log/status.pl: $@" if $@;
-   die "couldn't do $TopDir/log/status.pl: $!"    unless defined $ret;
-   die "couldn't run $TopDir/log/status.pl";
+if ( -f "$LogDir/status.pl" && !(my $ret = do "$LogDir/status.pl") ) {
+    if ( $@ ) {
+        print STDERR "couldn't parse $LogDir/status.pl: $@";
+    } elsif ( !defined($ret) ) {
+        print STDERR "couldn't do $LogDir/status.pl: $!";
+    } else {
+        print STDERR "couldn't run $LogDir/status.pl";
+    }
 }
 
 #
@@ -241,14 +248,16 @@ sub Main_Initialize
     umask($Conf{UmaskMode});
 
     #
-    # Check for another running process, check that PASSWD is set and
-    # verify executables are configured correctly.
+    # Check for another running process, verify executables are configured
+    # correctly and make sure $TopDir is on a file system that supports
+    # hardlinks.
     #
     if ( $Info{pid} ne "" && kill(0, $Info{pid}) ) {
         print(STDERR $bpc->timeStamp,
                  "Another BackupPC is running (pid $Info{pid}); quitting...\n");
         exit(1);
     }
+
     foreach my $progName ( qw(SmbClientPath NmbLookupPath PingPath DfPath
                               SendmailPath SshPath) ) {
         next if ( $Conf{$progName} eq "" || -x $Conf{$progName} );
@@ -258,12 +267,35 @@ sub Main_Initialize
         exit(1);
     }
 
+    if ( !$bpc->HardlinkTest("$TopDir/pc", "$TopDir/cpool") ) {
+        print(STDERR $bpc->timeStamp, "Can't create a test hardlink between a file"
+                   . " in $TopDir/pc and $TopDir/cpool.  Either these are different"
+                   . " file systems, or this file system doesn't support hardlinks,"
+                   . " or these directories don't exist, or there is a permissions"
+                   . " problem, or the file system is out of inodes or full.  Use"
+                   . " df, df -i, and ls -ld to check each of these possibilities."  
+                   . " Quitting...\n");
+        exit(1);
+    }
+
     if ( $opts{d} ) {
         #
-        # daemonize by forking
+        # daemonize by forking; more robust method per:
+        #       http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=301057
         #
-        defined(my $pid = fork) or die "Can't fork: $!";
+        my $pid;
+        defined($pid = fork) or die("Can't fork: $!");
+        exit if ( $pid );   # parent exits
+
+        POSIX::setsid();
+        defined($pid = fork) or die("Can't fork: $!");
         exit if $pid;   # parent exits
+
+        chdir ("/") or die("Cannot chdir to /: $!\n");
+        close(STDIN);
+        open(STDIN , ">/dev/null") or die("Cannot open /dev/null as stdin\n");
+        # STDOUT and STDERR are handled in LogFileOpen() right below,
+        # otherwise we would have to reopen them too.
     }
 
     #
@@ -331,9 +363,11 @@ sub Main_Initialize
     # Write out our initial status and save our PID
     #
     StatusWrite();
-    if ( open(PID, ">", "$TopDir/log/BackupPC.pid") ) {
+    unlink("$LogDir/BackupPC.pid");
+    if ( open(PID, ">", "$LogDir/BackupPC.pid") ) {
         print(PID $$);
         close(PID);
+        chmod(0444, "$LogDir/BackupPC.pid");
     }
 
     #
@@ -374,8 +408,7 @@ sub Main_TryToRun_nightly
            });
        $CmdQueueOn{$bpc->trashJob} = 1;
     }
-    if ( keys(%Jobs) == $trashCleanRunning && $RunNightlyWhenIdle == 1 ) {
-
+    if ( $RunNightlyWhenIdle == 1 ) {
         #
         # Queue multiple nightly jobs based on the configuration
         #
@@ -403,9 +436,6 @@ sub Main_TryToRun_nightly
 
        #
        # Zero out the data we expect to get from BackupPC_nightly.
-       # In the future if we want to split BackupPC_nightly over
-       # more than one night we will only zero out the portion
-       # that we are running right now.
        #
        for my $p ( qw(pool cpool) ) {
            for ( my $i = $start ; $i < $end ; $i++ ) {
@@ -429,7 +459,7 @@ sub Main_TryToRun_nightly
 
        #
        # Now queue the $Conf{MaxBackupPCNightlyJobs} jobs.
-       # The granularity on start and end is now 0..256.
+       # The granularity on start and end is now 0..255.
        #
        $start *= 16;
        $end   *= 16;
@@ -446,7 +476,6 @@ sub Main_TryToRun_nightly
             $start = $start0 + int(($end - $start0)
                                  * ($i + 1) / $Conf{MaxBackupPCNightlyJobs});
             push(@$cmd, $start - 1);
-
             my $job = $bpc->adminJob($i);
             unshift(@CmdQueue, {
                     host    => $job,
@@ -457,7 +486,6 @@ sub Main_TryToRun_nightly
             $CmdQueueOn{$job} = 1;
         }
         $RunNightlyWhenIdle = 2;
-
     }
 }
 
@@ -505,6 +533,8 @@ sub Main_TryToRun_CmdQueue
         }
         if ( !$pid ) {
             setpgrp 0,0;
+            $ENV{BPC_REQUSER} = $req->{user};
+            POSIX::nice($Conf{CmdQueueNice}) if ( $Conf{CmdQueueNice} );
             exec(@$cmd);
             print(LOG $bpc->timeStamp, "can't exec @$cmd for $host\n");
             exit(0);
@@ -546,14 +576,77 @@ sub Main_TryToRun_CmdQueue
 sub Main_TryToRun_Bg_or_User_Queue
 {
     my($req, $host);
-    while ( $RunNightlyWhenIdle == 0 ) {
+    my(@deferUserQueue, @deferBgQueue);
+    my $du;
+
+    if ( time - $Info{DUlastValueTime} >= 600 ) {
+        #
+        # Update our notion of disk usage no more than
+        # once every 10 minutes
+        #
+        $du = $bpc->CheckFileSystemUsage($TopDir);
+        $Info{DUlastValue}     = $du;
+        $Info{DUlastValueTime} = time;
+    } else {
+        #
+        # if we recently checked it then just use the old value
+        #
+        $du = $Info{DUlastValue};
+    }
+    if ( $Info{DUDailyMaxReset} ) {
+        $Info{DUDailyMaxStartTime} = time;
+        $Info{DUDailyMaxReset}     = 0;
+        $Info{DUDailyMax}          = 0;
+    }
+    if ( $du > $Info{DUDailyMax} ) {
+        $Info{DUDailyMax}     = $du;
+        $Info{DUDailyMaxTime} = time;
+    }
+    if ( $du > $Conf{DfMaxUsagePct} ) {
+        my @bgQueue = @BgQueue;
+        my $nSkip = 0;
+
+        #
+        # When the disk is too full, only run backups that will
+        # do expires, not regular backups
+        #
+        @BgQueue = ();
+        foreach $req ( @bgQueue ) {
+            if ( $req->{dumpExpire} ) {
+                unshift(@BgQueue, $req);
+            } else {
+                $BgQueueOn{$req->{host}} = 0;
+                $nSkip++;
+            }
+        }
+        if ( $nSkip ) {
+            print(LOG $bpc->timeStamp,
+                       "Disk too full ($du%); skipped $nSkip hosts\n");
+            $Info{DUDailySkipHostCnt} += $nSkip;
+        }
+    }
+
+    #
+    # Run background jobs anytime.  Previously they were locked out
+    # when BackupPC_nightly was running or pending with this
+    # condition on the while loop:
+    #
+    #    while ( $RunNightlyWhenIdle == 0 )
+    #
+    while ( 1 ) {
         local(*FH);
-        my(@args, @deferUserQueue, @deferBgQueue, $progName, $type);
+        my(@args, $progName, $type);
         my $nJobs = keys(%Jobs);
         #
         # CmdJob and trashClean don't count towards MaxBackups / MaxUserBackups
         #
-        $nJobs -= $BackupPCNightlyJobs if ( $CmdJob ne "" );
+        if ( $CmdJob ne "" ) {
+            if ( $BackupPCNightlyJobs ) {
+                $nJobs -= $BackupPCNightlyJobs;
+            } else {
+                $nJobs--;
+            }
+        }
         $nJobs-- if ( defined($Jobs{$bpc->trashJob} ) );
         if ( $nJobs < $Conf{MaxBackups} + $Conf{MaxUserBackups}
                         && @UserQueue > 0 ) {
@@ -562,60 +655,26 @@ sub Main_TryToRun_Bg_or_User_Queue
                 push(@deferUserQueue, $req);
                 next;
             }
-            push(@args, $req->{doFull} ? "-f" : "-i")
-                            if (( !$req->{restore} ) && ( !$req->{archive} ));
             $UserQueueOn{$req->{host}} = 0;
         } elsif ( $nJobs < $Conf{MaxBackups}
                         && (@CmdQueue + $nJobs)
                                 <= $Conf{MaxBackups} + $Conf{MaxPendingCmds}
                         && @BgQueue > 0 ) {
-            my $du;
-            if ( time - $Info{DUlastValueTime} >= 60 ) {
-                #
-                # Update our notion of disk usage no more than
-                # once every minute
-                #
-                $du = $bpc->CheckFileSystemUsage($TopDir);
-                $Info{DUlastValue}     = $du;
-                $Info{DUlastValueTime} = time;
-            } else {
-                #
-                # if we recently checked it then just use the old value
-                #
-                $du = $Info{DUlastValue};
-            }
-           if ( $Info{DUDailyMaxReset} ) {
-               $Info{DUDailyMaxStartTime} = time;
-               $Info{DUDailyMaxReset}     = 0;
-               $Info{DUDailyMax}          = 0;
-           }
-           if ( $du > $Info{DUDailyMax} ) {
-               $Info{DUDailyMax}     = $du;
-               $Info{DUDailyMaxTime} = time;
-           }
-            if ( $du > $Conf{DfMaxUsagePct} ) {
-                my $nSkip = @BgQueue + @deferBgQueue;
-                print(LOG $bpc->timeStamp,
-                           "Disk too full ($du%%); skipping $nSkip hosts\n");
-               $Info{DUDailySkipHostCnt} += $nSkip;
-                @BgQueue = ();
-                @deferBgQueue = ();
-                %BgQueueOn = ();
-                next;
-            }
             $req = pop(@BgQueue);
             if ( defined($Jobs{$req->{host}}) ) {
-                push(@deferBgQueue, $req);
+                #
+                # Job is currently running for this host; save it for later
+                #
+                unshift(@deferBgQueue, $req);
                 next;
             }
             $BgQueueOn{$req->{host}} = 0;
         } else {
-            while ( @deferBgQueue ) {
-                push(@BgQueue, pop(@deferBgQueue));
-            }
-            while ( @deferUserQueue ) {
-                push(@UserQueue, pop(@deferUserQueue));
-            }
+            #
+            # Restore the deferred jobs
+            #
+            @BgQueue   = (@BgQueue,   @deferBgQueue);
+            @UserQueue = (@UserQueue, @deferUserQueue);
             last;
         }
         $host = $req->{host};
@@ -631,7 +690,11 @@ sub Main_TryToRun_Bg_or_User_Queue
         } else {
             $progName = "BackupPC_dump";
             $type     = "backup";
-            push(@args, "-d") if ( $req->{dhcp} );
+            push(@args, "-I") if ( $req->{backupType} eq "autoIncr" );
+            push(@args, "-F") if ( $req->{backupType} eq "autoFull" );
+            push(@args, "-i") if ( $req->{backupType} eq "doIncr" );
+            push(@args, "-f") if ( $req->{backupType} eq "doFull" );
+            push(@args, "-d") if ( $req->{backupType} eq "dhcpPoll" );
             push(@args, "-e") if ( $req->{dumpExpire} );
             push(@args, $host);
         }
@@ -770,39 +833,47 @@ sub Main_Check_Timeout
         $Info{DUDailySkipHostCntPrev} = $Info{DUDailySkipHostCnt};
         $Info{DUDailySkipHostCnt}     = 0;
         my $lastLog = $Conf{MaxOldLogFiles} - 1;
-        if ( -f "$TopDir/log/LOG.$lastLog" ) {
+        if ( -f "$LogDir/LOG.$lastLog" ) {
             print(LOG $bpc->timeStamp,
-                       "Removing $TopDir/log/LOG.$lastLog\n");
-            unlink("$TopDir/log/LOG.$lastLog");
+                       "Removing $LogDir/LOG.$lastLog\n");
+            unlink("$LogDir/LOG.$lastLog");
         }
-        if ( -f "$TopDir/log/LOG.$lastLog.z" ) {
+        if ( -f "$LogDir/LOG.$lastLog.z" ) {
             print(LOG $bpc->timeStamp,
-                       "Removing $TopDir/log/LOG.$lastLog.z\n");
-            unlink("$TopDir/log/LOG.$lastLog.z");
+                       "Removing $LogDir/LOG.$lastLog.z\n");
+            unlink("$LogDir/LOG.$lastLog.z");
         }
         print(LOG $bpc->timeStamp, "Aging LOG files, LOG -> LOG.0 -> "
                    . "LOG.1 -> ... -> LOG.$lastLog\n");
+       close(STDERR);          # dup of LOG
+       close(STDOUT);          # dup of LOG
         close(LOG);
         for ( my $i = $lastLog - 1 ; $i >= 0 ; $i-- ) {
             my $j = $i + 1;
-            rename("$TopDir/log/LOG.$i", "$TopDir/log/LOG.$j")
-                            if ( -f "$TopDir/log/LOG.$i" );
-            rename("$TopDir/log/LOG.$i.z", "$TopDir/log/LOG.$j.z")
-                            if ( -f "$TopDir/log/LOG.$i.z" );
+            rename("$LogDir/LOG.$i", "$LogDir/LOG.$j")
+                            if ( -f "$LogDir/LOG.$i" );
+            rename("$LogDir/LOG.$i.z", "$LogDir/LOG.$j.z")
+                            if ( -f "$LogDir/LOG.$i.z" );
         }
        #
        # Compress the log file LOG -> LOG.0.z (if enabled).
        # Otherwise, just rename LOG -> LOG.0.
        #
-       BackupPC::FileZIO->compressCopy("$TopDir/log/LOG",
-                                        "$TopDir/log/LOG.0.z",
-                                        "$TopDir/log/LOG.0",
+       BackupPC::FileZIO->compressCopy("$LogDir/LOG",
+                                        "$LogDir/LOG.0.z",
+                                        "$LogDir/LOG.0",
                                         $Conf{CompressLevel}, 1);
         LogFileOpen();
         #
-        # Remember to run nightly script after current jobs are done
+        # Remember to run the nightly script when the next CmdQueue
+        # job is done.
         #
-        $RunNightlyWhenIdle = 1;
+        if ( $RunNightlyWhenIdle == 2 ) {
+            print(LOG $bpc->timeStamp, "BackupPC_nightly is still running after 24 hours!!"
+                                     . "  You should adjust the config settings; Skipping this run\n");
+        } else {
+            $RunNightlyWhenIdle = 1;
+        }
     }
     #
     # Write out the current status and then queue all the PCs
@@ -892,6 +963,8 @@ sub Main_Check_Job_Messages
                                 if ( $Status{$host}{dhcpCheckCnt} > 0 );
             } elsif ( $mesg =~ /^xferPids (.*)/ ) {
                 $Jobs{$host}{xferPid} = $1;
+            } elsif ( $mesg =~ /^completionPercent (.*)/ ) {
+                $Jobs{$host}{completionPercent} = $1;
             } elsif ( $mesg =~ /^started_restore/ ) {
                 $Jobs{$host}{type}    = "restore";
                 print(LOG $bpc->timeStamp,
@@ -920,6 +993,7 @@ sub Main_Check_Job_Messages
                 delete($Status{$host}{error});
                 delete($Status{$host}{errorTime});
                 $Status{$host}{endTime}   = time;
+                $Status{$host}{lastGoodBackupTime} = time;
             } elsif ( $mesg =~ /^backups disabled/ ) {
                 print(LOG $bpc->timeStamp,
                            "Ignoring old backup error on $host\n");
@@ -1022,14 +1096,16 @@ sub Main_Check_Job_Messages
                 $Info{pool}{$f[0]}[$chunk]{FileCntRename} += $f[9];
                 $Info{pool}{$f[0]}[$chunk]{FileLinkMax}    = $f[10]
                         if ( $Info{pool}{$f[0]}[$chunk]{FileLinkMax} < $f[10] );
+                $Info{pool}{$f[0]}[$chunk]{FileLinkTotal} += $f[11];
                 $Info{pool}{$f[0]}[$chunk]{Time}           = time;
             } elsif ( $mesg =~ /^BackupPC_nightly lock_off/ ) {
                $BackupPCNightlyLock--;
                 if ( $BackupPCNightlyLock == 0 ) {
                     #
                     # This means the last BackupPC_nightly is done with
-                   # the pool clean, so it's to start running regular
-                   # backups again.
+                   # the pool clean, so it's ok to start running regular
+                   # backups again.  But starting in 3.0 regular jobs
+                    # are decoupled from BackupPC_nightly.
                     #
                     $RunNightlyWhenIdle = 0;
                }
@@ -1059,6 +1135,9 @@ sub Main_Check_Job_Messages
                    #print(LOG $bpc->timeStamp, "BackupPC_nightly done; now"
                    #         . " have $BackupPCNightlyJobs running\n");
                     if ( $BackupPCNightlyJobs <= 0 ) {
+                        #
+                        # Last BackupPC_nightly has finished
+                        #
                         $BackupPCNightlyJobs = 0;
                         $RunNightlyWhenIdle = 0;
                         $CmdJob = "";
@@ -1210,7 +1289,7 @@ sub Main_Check_Client_Messages
                 last;
             }
             $Clients{$client}{mesgCnt}++;
-            $cmd = $2;
+            $cmd = decode_utf8($2);
             if ( $cmd =~ /^stop (\S+)\s+(\S+)\s+(\S*)/ ) {
                 $host = $1;
                 my $user = $2;
@@ -1271,45 +1350,65 @@ sub Main_Check_Client_Messages
             } elsif ( $cmd =~ /^BackupPC_nightly run$/ ) {
                 $RunNightlyWhenIdle = 1;
             } elsif ( $cmd =~ /^backup (\S+)\s+(\S+)\s+(\S+)\s+(\S+)/ ) {
-                my $hostIP = $1;
-                $host      = $2;
-                my $user   = $3;
-                my $doFull = $4;
-               $host      = $bpc->uriUnesc($host);
-               $hostIP    = $bpc->uriUnesc($hostIP);
-                if ( !defined($Status{$host}) ) {
+                my $hostIP     = $1;
+                $host          = $2;
+                my $user       = $3;
+                my $backupType = $4;
+               $host          = $bpc->uriUnesc($host);
+               $hostIP        = $bpc->uriUnesc($hostIP);
+                if ( !defined($Hosts->{$host}) ) {
                     print(LOG $bpc->timeStamp,
                                "User $user requested backup of unknown host"
                              . " $host\n");
                     $reply = "error: unknown host $host";
-                } elsif ( defined($Jobs{$host})
-                                && $Jobs{$host}{type} ne "restore" ) {
-                    print(LOG $bpc->timeStamp,
-                               "User $user requested backup of $host,"
-                             . " but one is currently running\n");
-                    $reply = "error: backup of $host is already running";
                 } else {
-                    print(LOG $bpc->timeStamp,
-                               "User $user requested backup of $host"
-                             . " ($hostIP)\n");
-                    if ( $BgQueueOn{$hostIP} ) {
-                        @BgQueue = grep($_->{host} ne $hostIP, @BgQueue);
-                        $BgQueueOn{$hostIP} = 0;
-                    }
-                    if ( $UserQueueOn{$hostIP} ) {
-                        @UserQueue = grep($_->{host} ne $hostIP, @UserQueue);
-                        $UserQueueOn{$hostIP} = 0;
+                    #
+                    # Handle numeric backupType for backward compatibility
+                    # (technically -1 is a new feature for auto)
+                    #
+                    $backupType = 'auto'   if ( $backupType eq '-1' );
+                    $backupType = 'doIncr' if ( $backupType eq '0' );
+                    $backupType = 'doFull' if ( $backupType eq '1' );
+                    if ( $backupType !~ /^doIncr|doFull|autoIncr|autoFull|auto$/i ) {
+                        $reply = "error: unknown backup type $backupType";
+                    } else {
+                        print(LOG $bpc->timeStamp,
+                                   "User $user requested backup of $host"
+                                 . " ($hostIP)\n");
+                        if ( $BgQueueOn{$hostIP} ) {
+                            @BgQueue = grep($_->{host} ne $hostIP, @BgQueue);
+                            $BgQueueOn{$hostIP} = 0;
+                        }
+                        if ( $UserQueueOn{$hostIP} ) {
+                            @UserQueue = grep($_->{host} ne $hostIP, @UserQueue);
+                            $UserQueueOn{$hostIP} = 0;
+                        }
+                        my $status = QueueOnePC($host, $hostIP, $user, 'user', $backupType);
+                        if ( $status == 0 ) {
+                            $reply = "ok: requested backup of $host ($backupType)";
+                        } elsif ( $status == 1 ) {
+                            #should never see this we just dequeued it
+                            $reply = "warning: $host was already queued."
+                                   . " Ignoring this request";
+                        } elsif ( $status == 2 ) {
+                            print(LOG $bpc->timeStamp,
+                                   "Disk too full ($Info{DUlastValue}%)."
+                                 . " Not queueing backup of $host\n");
+                            $reply = "error: disk too full ($Info{DUlastValue}%)";
+                            $Info{DUDailySkipHostCnt}++;
+                        } elsif ( $status == 3 ) {
+                            # should never reach this because
+                            # it's set to "user" above
+                            $reply = "error: unknown queue name";
+                        } else {
+                            $reply = "error: unknown queue status $status";
+                            if ( $BgQueueOn{$hostIP} || $UserQueueOn{$hostIP} ) {
+                                $reply .= ".  Host is queued.";
+                            } else {
+                                $reply .= ".  Host is not queued.";
+                            }
+                        }
                     }
-                    unshift(@UserQueue, {
-                                host    => $hostIP,
-                                user    => $user,
-                                reqTime => time,
-                                doFull  => $doFull,
-                                userReq => 1,
-                                dhcp    => $hostIP eq $host ? 0 : 1,
-                        });
-                    $UserQueueOn{$hostIP} = 1;
-                    $reply = "ok: requested backup of $host";
                 }
             } elsif ( $cmd =~ /^archive (\S+)\s+(\S+)\s+(\S+)/ ) {
                 my $user         = $1;
@@ -1330,7 +1429,7 @@ sub Main_Check_Client_Messages
                     } else {
                         unshift(@UserQueue, {
                                 host    => $host,
-                                hostIP  => $user,
+                                user    => $user,
                                 reqFileName => $reqFileName,
                                 reqTime => time,
                                 dhcp    => 0,
@@ -1348,7 +1447,7 @@ sub Main_Check_Client_Messages
                 my $reqFileName = $4;
                $host      = $bpc->uriUnesc($host);
                $hostIP    = $bpc->uriUnesc($hostIP);
-                if ( !defined($Status{$host}) ) {
+                if ( !defined($Hosts->{$host}) ) {
                     print(LOG $bpc->timeStamp,
                                "User $user requested restore to unknown host"
                              . " $host");
@@ -1517,7 +1616,7 @@ sub Main_Check_Client_Messages
 ###########################################################################
 
 #
-# Write the current status to $TopDir/log/status.pl
+# Write the current status to $LogDir/status.pl
 #
 sub StatusWrite
 {
@@ -1525,70 +1624,145 @@ sub StatusWrite
              [  \%Info, \%Status],
              [qw(*Info   *Status)]);
     $dump->Indent(1);
-    if ( open(STATUS, ">", "$TopDir/log/status.pl") ) {
-        print(STATUS $dump->Dump);
-        close(STATUS);
-    }
+    my $text = $dump->Dump;
+    $bpc->{storage}->TextFileWrite("$LogDir/status.pl", $text);
 }
 
 #
-# Queue all the hosts for backup.  This means queuing all the fixed
-# ip hosts and all the dhcp address ranges.  We also additionally
-# queue the dhcp hosts with a -e flag to check for expired dumps.
+# Compare function for host sort.  Hosts with errors go first,
+# sorted with the oldest errors first.  The remaining hosts
+# are sorted so that those with the oldest backups go first.
 #
-sub QueueAllPCs
+sub HostSortCompare
 {
-    foreach my $host ( sort(keys(%$Hosts)) ) {
-        delete($Status{$host}{backoffTime})
-                if ( defined($Status{$host}{backoffTime})
+    #
+    # Hosts with errors go before hosts without errors
+    #
+    return -1 if ( $Status{$a}{error} ne "" && $Status{$b}{error} eq "" );
+
+    #
+    # Hosts with no errors go after hosts with errors
+    #
+    return  1 if ( $Status{$a}{error} eq "" && $Status{$b}{error} ne "" );
+
+    #
+    # hosts with the older last good backups sort earlier
+    #
+    my $r = $Status{$a}{lastGoodBackupTime} <=> $Status{$b}{lastGoodBackupTime};
+    return $r if ( $r );
+
+    #
+    # Finally, just sort based on host name
+    #
+    return $a cmp $b;
+}
+
+#
+# Attempt to queue a host.
+# Returns 0 on success; 1 if host is already queued;
+#         2 if host was skipped; 3 on invalid queue name
+#
+# $host       is the client's host name
+# $hostIP     is usually the client's host name too, or IP address
+#             if the user specified it in the manual backup command
+# $user       is the user name, or BackupPC by default
+# $queue      is which queue to use ("bg" by default)
+# $backupType is the backup type (doIncr|doFull|autoIncr|autoFull|auto|dhcpPoll)
+#
+# Note: starting in 3.2.0, the PC is queued even if it has a current
+# job running
+#
+sub QueueOnePC
+{
+    my($host, $hostIP, $user, $queue, $backupType) = @_;
+    my $retVal  = 0;
+    $user       = "BackupPC" if ( $user eq '' );
+    $queue      = "bg"   if ( $queue eq '' && $user eq 'BackupPC' );
+    $backupType = "auto" if ( $backupType eq '' );
+
+    delete($Status{$host}{backoffTime})
+            if ( defined($Status{$host}{backoffTime})
                   && $Status{$host}{backoffTime} < time );
-        next if ( defined($Jobs{$host})
-                || $BgQueueOn{$host}
-                || $UserQueueOn{$host}
-                || $CmdQueueOn{$host} ); 
-        if ( $Hosts->{$host}{dhcp} ) {
-            $Status{$host}{dhcpCheckCnt}++;
-            if ( $RunNightlyWhenIdle ) {
-                #
-                # Once per night queue a check for DHCP hosts that just
-                # checks for expired dumps.  We need to do this to handle
-                # the case when a DHCP host has not been on the network for
-                # a long time, and some of the old dumps need to be expired.
-                # Normally expiry checks are done by BackupPC_dump only
-                # after the DHCP hosts has been detected on the network.
-                #
-                unshift(@BgQueue,
-                    {host => $host, user => "BackupPC", reqTime => time,
-                     dhcp => 0, dumpExpire => 1});
-                $BgQueueOn{$host} = 1;
-            }
-        } else {
+    return 1 if ( $BgQueueOn{$host} || $UserQueueOn{$host} );
+    if ( $Hosts->{$host}{dhcp} ) {
+        $Status{$host}{dhcpCheckCnt}++;
+        if ( $RunNightlyWhenIdle ) {
             #
-            # this is a fixed ip host: queue it
+            # Once per night queue a check for DHCP hosts that just
+            # checks for expired dumps.  We need to do this to handle
+            # the case when a DHCP host has not been on the network for
+            # a long time, and some of the old dumps need to be expired.
+            # Normally expiry checks are done by BackupPC_dump only
+            # after the DHCP hosts has been detected on the network.
             #
             unshift(@BgQueue,
-                {host => $host, user => "BackupPC", reqTime => time,
-                 dhcp => $Hosts->{$host}{dhcp}});
+                {host => $hostIP, user => $user, reqTime => time,
+                 dhcp => 0, dumpExpire => 1});
             $BgQueueOn{$host} = 1;
         }
+    } else {
+        #
+        # this is a fixed ip host or DHCP ip address: queue it
+        #
+        if ( $Info{DUlastValue} > $Conf{DfMaxUsagePct} ) {
+            #
+            # Since we are out of disk space, instead of queuing
+            # a regular job, queue an expire check instead.  That
+            # way if the admin reduces the number of backups to
+            # keep then we will actually delete them.  Otherwise
+            # BackupPC_dump will never run since we have exceeded
+            # the limit.
+            #
+            $retVal = 2;
+            unshift(@BgQueue,
+                {host => $hostIP, user => $user, reqTime => time, dumpExpire => 1});
+            $BgQueueOn{$host} = 1;
+        } elsif( $queue eq 'bg' ) {
+            #
+            # Queue regular background backup
+            #
+            unshift(@BgQueue,
+                {host => $hostIP, user => $user, reqTime => time, backupType => $backupType});
+            $BgQueueOn{$host} = 1;
+        } elsif( $queue eq 'user' ) {
+            #
+            # Queue user backup
+            #
+            unshift(@UserQueue,
+                {host => $hostIP, user => $user, reqTime => time, backupType => $backupType});
+            $UserQueueOn{$host} = 1;
+        } else {
+            # unknown $queue type
+            $retVal = 3;
+        }
+    }
+
+    return $retVal;
+}
+
+#
+# Queue all the hosts for backup.  This means queuing all the fixed
+# ip hosts and all the dhcp address ranges.  We also additionally
+# queue the dhcp hosts with a -e flag to check for expired dumps.
+#
+sub QueueAllPCs
+{
+    my $nSkip = 0;
+
+    foreach my $host ( sort HostSortCompare keys(%$Hosts) ) {
+        $nSkip++ if ( QueueOnePC($host, $host, 'BackupPC', 'bg', 'auto') == 2 );
     }
     foreach my $dhcp ( @{$Conf{DHCPAddressRanges}} ) {
         for ( my $i = $dhcp->{first} ; $i <= $dhcp->{last} ; $i++ ) {
             my $ipAddr = "$dhcp->{ipAddrBase}.$i";
-            next if ( defined($Jobs{$ipAddr})
-                    || $BgQueueOn{$ipAddr}
-                    || $UserQueueOn{$ipAddr}
-                    || $CmdQueueOn{$ipAddr} ); 
-            #
-            # this is a potential dhcp ip address (we don't know the
-            # host name yet): queue it
-            #
-            unshift(@BgQueue,
-                {host => $ipAddr, user => "BackupPC", reqTime => time,
-                 dhcp => 1});
-            $BgQueueOn{$ipAddr} = 1;
+            $nSkip++ if ( QueueOnePC($ipAddr, $ipAddr, 'BackupPC', 'bg', 'dhcpPoll') == 2 );
         }
     }
+    if ( $nSkip ) {
+        print(LOG $bpc->timeStamp,
+               "Disk too full ($Info{DUlastValue}%); skipped $nSkip hosts\n");
+        $Info{DUDailySkipHostCnt} += $nSkip;
+    }
 }
 
 #
@@ -1672,7 +1846,7 @@ sub catch_signal
         close(LOG);
         LogFileOpen();
         print(LOG "Fatal error: unhandled signal $SigName\n");
-        unlink("$TopDir/log/BackupPC.pid");
+        unlink("$LogDir/BackupPC.pid");
         confess("Got new signal $SigName... quitting\n");
     } else {
        $SigName = shift;
@@ -1684,9 +1858,9 @@ sub catch_signal
 #
 sub LogFileOpen
 {
-    mkpath("$TopDir/log", 0, 0777) if ( !-d "$TopDir/log" );
-    open(LOG, ">>$TopDir/log/LOG")
-            || die("Can't create LOG file $TopDir/log/LOG");
+    mkpath($LogDir, 0, 0777) if ( !-d $LogDir );
+    open(LOG, ">>$LogDir/LOG")
+            || die("Can't create LOG file $LogDir/LOG");
     close(STDOUT);
     close(STDERR);
     open(STDOUT, ">&LOG");
@@ -1711,7 +1885,7 @@ sub ServerSocketInit
             print(LOG $bpc->timeStamp, "unix socket() failed: $!\n");
             exit(1);
         }
-        my $sockFile = "$TopDir/log/BackupPC.sock";
+        my $sockFile = "$LogDir/BackupPC.sock";
         unlink($sockFile);
         if ( !bind(SERVER_UNIX, sockaddr_un($sockFile)) ) {
             print(LOG $bpc->timeStamp, "unix bind() failed: $!\n");
@@ -1792,8 +1966,9 @@ sub ServerShutdown
         }
         %Jobs = ();
     }
+    delete($Info{pid});
     StatusWrite();
-    unlink("$TopDir/log/BackupPC.pid");
+    unlink("$LogDir/BackupPC.pid");
     exit(1);
 }