X-Git-Url: http://git.freeside.biz/gitweb/?a=blobdiff_plain;f=FS%2Fbin%2Ffreeside-torrus-srvderive;h=258385173fc2db9248b37fa3bca5fb12007df973;hb=5d19829b734c590e57a301833268be89713deb27;hp=ce47414a698b763af37be322cdf8a77dca572d35;hpb=2ea1bf50d3c8c08c208a412d9f468d9dfccb0b60;p=freeside.git diff --git a/FS/bin/freeside-torrus-srvderive b/FS/bin/freeside-torrus-srvderive index ce47414a6..258385173 100644 --- a/FS/bin/freeside-torrus-srvderive +++ b/FS/bin/freeside-torrus-srvderive @@ -1,14 +1,20 @@ #!/usr/bin/perl -w use strict; +use POSIX qw( :sys_wait_h ); +use Sys::SigAction qw( set_sig_handler ); use Date::Parse; use Date::Format; +use IPC::Run qw( run ); #timeout ); use FS::Daemon ':all'; #daemonize1 drop_root daemonize2 myexit logfile sig* -use FS::UID qw( adminsuidsetup dbh driver_name ); +use FS::UID qw( adminsuidsetup forksuidsetup dbh driver_name ); use FS::Record qw( qsearch str2time_sql str2time_sql_closing concat_sql ); use FS::torrus_srvderive; -our $DEBUG = 3; +our $DEBUG = 2; +our $max_kids = 16; +our $search_timeout = 2*60*60; #60*60; #30*60; # 15*60 5*60; #$torrus_srvderive->last_srv_date ? 5*60 : 15*60); +our %kids; my $user = shift or die &usage; $FS::Daemon::PID_NEWSTYLE = 1; @@ -51,72 +57,249 @@ my $sql = " AND srvexport.srv_date = other.srv_date AND ABS( $_date - $other_date ) <= 60 ) +"; + +my $orderlimit = " ORDER BY srv_date, srv_time - LIMIT 100 + LIMIT 200 "; +if ( driver_name =~ /mysql/i ) { + dbh->do('SET SESSION TRANSACTION ISOLATION LEVEL READ UNCOMMITTED'); + dbh->commit or die dbh->errstr; +} + +our $kids = 0; + +#MAIN: while (1) { while (1) { my $found = 0; + #SERVICEID: foreach my $torrus_srvderive ( qsearch('torrus_srvderive', {}) ) { foreach my $torrus_srvderive ( qsearch('torrus_srvderive', {}) ) { - my $serviceid = $torrus_srvderive->serviceid; + &reap_kids; + if ( $kids >= $max_kids ) { + sleep 5; + myexit() if sigterm() || sigint(); + redo; + } + + defined( my $pid = fork ) or do { + #warn "WARNING: can't fork: $!\n"; + #next; #don't increment the kid counter + die "can't fork: $!\n"; + }; - warn $sql if $DEBUG > 2; - my $sth = dbh->prepare($sql) or die $DBI::errstr; #better recovery? - $sth->execute($serviceid, $serviceid) or die $sth->errstr; - my $prev = 0; - while ( my $row = $sth->fetchrow_arrayref ) { - last if sigterm() || sigint(); + if ( $pid ) { + $kids++; + $kids{$pid} = 1; + } else { #kid time - my( $srv_date, $srv_time ) = @$row; - my $cur = str2time( "$srv_date $srv_time" ); - next if $cur-$prev <= 60; - last if time - $cur <= 300; + #get new db handle + $FS::UID::dbh->{InactiveDestroy} = 1; - warn "no $serviceid for $srv_date $srv_time; adding\n" - if $DEBUG; - $found++; + forksuidsetup($user); + + my $serviceid = $torrus_srvderive->serviceid; my @serviceids = $torrus_srvderive->component_serviceids; + exit unless @serviceids; #don't try to search for empty virtual ports + + #nonlocking select statements; rows in this table never change + if ( driver_name =~ /mysql/i ) { + dbh->do('SET SESSION TRANSACTION ISOLATION LEVEL READ UNCOMMITTED'); + dbh->commit or die dbh->errstr; + } + my @in = (); for my $dir ('_IN', '_OUT') { + push @in, map dbh->quote("$_$dir"), @serviceids; + } + my $in = join(',', @in); + + if ( ! $torrus_srvderive->last_srv_date ) { + warn "finding initial last_srv_date for $serviceid\n" if $DEBUG; + my $dsql = "SELECT srv_date FROM srvexport WHERE serviceid IN ($in) + ORDER BY srv_date LIMIT 1"; + my $dsth = dbh->prepare($dsql) or die $DBI::errstr; + $dsth->execute or die $dsth->errstr; + my $date = $dsth->fetchrow_arrayref->[0]; + if ( $date ) { + warn "found initial last_srv_date of $date; updating $serviceid\n" + if $DEBUG; + $torrus_srvderive->last_srv_date($date); + my $error = $torrus_srvderive->replace; + die $error if $error; + } else { + warn "no initial last_srv_date for $serviceid; skipping\n" if $DEBUG; + exit; + } + } - my $sin = join(',', map dbh->quote("$_$dir"), @serviceids); - - my $sum = " - SELECT COALESCE(SUM(value),0) FROM srvexport AS other - WHERE other.serviceid IN ($sin) - AND ABS( $cur - $other_date ) <= 60 - "; - - my $isql = " - INSERT INTO srvexport ( srv_date, srv_time, serviceid, value, intvl ) - VALUES ( ?, ?, ?, ($sum), ? ) - "; - my @param = ( time2str('%Y-%m-%d', $cur), #srv_date - time2str('%X', $cur), #srv_time - "$serviceid$dir", - 300, #intvl ... + #alternate (hopefully more efficient) path: if we're more than a day + #behind, call native torrus srvderive + my $last_srv_date = $torrus_srvderive->last_srv_date; + my $currdate = time2str('%Y-%m-%d', time); + if ( $last_srv_date ne $currdate + && time - str2time($last_srv_date) > (24+6)*60*60 #6 hour window for + ) { #yesterday + warn "serviceid $serviceid last_srv_date != $currdate\n" if $DEBUG; + + #delete records from last_srv_date + my $dsql = + 'DELETE FROM srvexport WHERE srv_date = ? AND serviceid IN(?,?)'; + my $dsth = dbh->prepare($dsql) or die $DBI::errstr; + $dsth->execute($last_srv_date, $serviceid.'_IN', $serviceid.'_OUT') + or die $dsth->errstr; + dbh->commit or die dbh->errstr; + + foreach my $dir ('_IN','_OUT') { + + #XX TORRUS_PATH + my @cmd = ( '/usr/local/bin/torrus', 'srvderive', + "--start=$last_srv_date", "--end=$currdate", + "--out=$serviceid$dir", "--func=SUM", + map "$_$dir", @serviceids ); - warn $isql. ' with param '. join(',',@param). "\n" - if $DEBUG > 1; - my $isth = dbh->prepare($isql) or die $DBI::errstr; #better recovery? - $isth->execute( @param ) or die $isth->errstr; - dbh->commit or die dbh->errstr; - + warn "calling native srvderive: ". join(' ', @cmd). "\n" if $DEBUG; + run( \@cmd ); #verbose? timeout? + + } + warn "native srvderive done" if $DEBUG; + + $torrus_srvderive->last_srv_date($currdate); + my $error = $torrus_srvderive->replace; + die $error if $error; + + #don't bother trying today's immediately + warn "done with $serviceid for now\n" if $DEBUG; + exit; #end-of-kid + } - $prev = $cur; - } + my $ssql = " + $sql AND EXISTS ( + SELECT 1 FROM srvexport AS other + WHERE other.serviceid IN ($in) + AND srvexport.srv_date = other.srv_date + AND ABS( $_date - $other_date ) <= 60 + ) + "; + + $ssql .= " AND srv_date >= '". $torrus_srvderive->last_srv_date. "' " + if $torrus_srvderive->last_srv_date; + + $ssql .= $orderlimit; + + warn "searching for times to add $serviceid\n" if $DEBUG; + warn $ssql if $DEBUG > 2; + my $sth = dbh->prepare($ssql) or die $DBI::errstr; #better recovery here? + + eval { + my $h = set_sig_handler( 'ALRM', sub { die "_timeout\n"; } ); + alarm($search_timeout); + $sth->execute($serviceid, $serviceid) or die $sth->errstr; + alarm(0); + }; + alarm(0); + + if ( $@ && $@ eq "_timeout\n" ) { + #warn "search timed out; reconnecting and restarting\n"; + warn "search timed out\n"; + dbh->clone()->do("KILL QUERY ". dbh->{"mysql_thread_id"}) + if driver_name =~ /mysql/i; + dbh->rollback; #or die dbh->errstr; + #adminsuidsetup($user); + #next SERVICEID; #MAIN; + exit; + } elsif ( $@ ) { + die $@; + } - } + warn "search for $serviceid finished; checking results\n" if $DEBUG; + + my $prev = 0; + while ( my $row = $sth->fetchrow_arrayref ) { + last if sigterm() || sigint(); + + my( $srv_date, $srv_time ) = @$row; + my $cur = str2time( "$srv_date $srv_time" ); + next if $cur-$prev <= 60; + last if time - $cur <= 300; + + warn "no $serviceid for $srv_date $srv_time; adding\n" + if $DEBUG; + $found++; + + for my $dir ('_IN', '_OUT') { + + my $sin = join(',', map dbh->quote("$_$dir"), @serviceids); + + my $srv_date = time2str('%Y-%m-%d', $cur); + + my $sum = " + SELECT COALESCE(SUM(value),0) FROM srvexport AS other + WHERE other.serviceid IN ($sin) + AND other.srv_date = '$srv_date' + AND ABS( $cur - $other_date ) <= 60 + "; + my $ssth = dbh->prepare($sum) or die $DBI::errstr; + $ssth->execute or die $ssth->errstr; #better recovery? + my $value = $ssth->fetchrow_arrayref->[0]; + + my $isql = " + INSERT INTO srvexport (srv_date, srv_time, serviceid, value, intvl) + VALUES ( ?, ?, ?, ?, ? ) + "; + my @param = ( $srv_date, + time2str('%X', $cur), #srv_time + "$serviceid$dir", #serviceid + $value, + 300, #intvl ... + ); + warn $isql. ' with param '. join(',',@param). "\n" + if $DEBUG > 2; + + my $isth = dbh->prepare($isql) or die $DBI::errstr; #better recovery? + + #stupid mysql deadlocks all the time on insert, so we need to recover + unless ( $isth->execute(@param) ) { + #warn "Error inserting data for $serviceid$dir (restarting): ". + # $isth->errstr; + warn "Error inserting data for $serviceid$dir: ". $isth->errstr; + dbh->rollback; #or die dbh->errstr; + #sleep 5; + #next SERVICEID; #MAIN; + exit; + } + + } + + if ( $srv_date ne $torrus_srvderive->last_srv_date ) { + warn "updating last_srv_date of $serviceid to $srv_date\n" if $DEBUG; + $torrus_srvderive->last_srv_date($srv_date); + my $error = $torrus_srvderive->replace; + die $error if $error; + } + dbh->commit or die dbh->errstr; + + $prev = $cur; + } + warn "done with $serviceid\n" if $DEBUG; + + exit; + #end-of-kid + } + } #foreach my $torrus_srvderive + dbh->commit or die dbh->errstr; myexit() if sigterm() || sigint(); - sleep 60 unless $found; + warn "restarting main loop\n" if $DEBUG > 1; + #sleep 60 unless $found; } sub _shouldrun { @@ -128,6 +311,15 @@ sub usage { die "Usage:\n\n freeside-cdrrewrited user\n"; } +sub reap_kids { + foreach my $pid ( keys %kids ) { + my $kid = waitpid($pid, WNOHANG); + if ( $kid > 0 ) { + $kids--; + delete $kids{$kid}; + } + } +} =head1 NAME