X-Git-Url: http://git.freeside.biz/gitweb/?p=freeside.git;a=blobdiff_plain;f=FS%2Fbin%2Ffreeside-torrus-srvderive;h=5ad23aba45cee579194db88de85a862b9af587a2;hp=23d2e44a9a257740c857e907b513cb1a836d42f0;hb=6de06472ab43534bd889e531ae060bbd4c935518;hpb=f9d08a9be9e03f985de40b56b2b3b8b37d1c49bf diff --git a/FS/bin/freeside-torrus-srvderive b/FS/bin/freeside-torrus-srvderive index 23d2e44a9..5ad23aba4 100644 --- a/FS/bin/freeside-torrus-srvderive +++ b/FS/bin/freeside-torrus-srvderive @@ -5,14 +5,14 @@ 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 forksuidsetup dbh driver_name ); -use FS::Record qw( qsearch str2time_sql str2time_sql_closing concat_sql ); +use FS::Record qw( qsearch ); use FS::torrus_srvderive; our $DEBUG = 2; -our $max_kids = 6; -our $search_timeout = 60*60; #30*60; # 15*60 5*60; #$torrus_srvderive->last_srv_date ? 5*60 : 15*60); +our $max_kids = 4; our %kids; my $user = shift or die &usage; @@ -34,47 +34,11 @@ die "not running: network_monitoring_system not Torrus_Internal\n" #-- -my $str2time = str2time_sql(); -my $c = str2time_sql_closing(); - -my $_date = concat_sql([ 'srvexport.srv_date', "' '", 'srvexport.srv_time' ]); -$_date = "CAST( $_date AS TIMESTAMP )" if driver_name =~ /^Pg/i; -$_date = str2time_sql. $_date. str2time_sql_closing; - -my $other_date = concat_sql([ 'other.srv_date', "' '", 'other.srv_time' ]); -$other_date = "CAST( $other_date AS TIMESTAMP )" if driver_name =~ /^Pg/i; -$other_date = str2time_sql. $other_date. str2time_sql_closing; - -my $in = concat_sql([ '?', "'_IN'" ]); -my $out = concat_sql([ '?', "'_OUT'" ]); - -my $sql = " - SELECT DISTINCT srv_date, srv_time FROM srvexport - WHERE NOT EXISTS ( - SELECT 1 FROM srvexport AS other - WHERE other.serviceid IN ( $in, $out ) - AND srvexport.srv_date = other.srv_date - AND ABS( $_date - $other_date ) <= 60 - ) -"; - -my $orderlimit = " - ORDER BY srv_date, srv_time - LIMIT 50 -"; - -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', {}) ) { @@ -106,12 +70,6 @@ while (1) { 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; @@ -124,7 +82,8 @@ while (1) { 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]; + my $row = $dsth->fetchrow_arrayref; + my $date = (defined($row) && ref($row)) ? $row->[0] : ''; if ( $date ) { warn "found initial last_srv_date of $date; updating $serviceid\n" if $DEBUG; @@ -137,126 +96,66 @@ while (1) { } } - 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 $@; - } + #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)*60*60 + ) { + 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 $DBI::errstr; + warn "$dsql $last_srv_date ${serviceid}_IN ${serviceid}_OUT\n" + if $DEBUG; + $dsth->execute($last_srv_date, $serviceid.'_IN', $serviceid.'_OUT') + or die $dsth->errstr; + dbh->commit or die dbh->errstr; - warn "search for $serviceid finished; checking results\n" if $DEBUG; + foreach my $dir ('_IN','_OUT') { - my $prev = 0; - while ( my $row = $sth->fetchrow_arrayref ) { - last if sigterm() || sigint(); + #XX TORRUS_PATH + my @cmd = ( '/usr/local/bin/torrus', 'srvderive', + "--start=$last_srv_date", "--end=$currdate", + "--out=$serviceid$dir", "--func=SUM", + map "$_$dir", @serviceids + ); - my( $srv_date, $srv_time ) = @$row; - my $cur = str2time( "$srv_date $srv_time" ); - next if $cur-$prev <= 60; - last if time - $cur <= 300; + warn "calling native srvderive: ". join(' ', @cmd). "\n" if $DEBUG; + my $out_and_err = ''; + run( \@cmd, '>&', \$out_and_err ) #verbose? timeout? + or die "error calling native srvderive: $out_and_err\n"; - 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; - } - } + warn "native srvderive done" if $DEBUG; - 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; + $torrus_srvderive->last_srv_date($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; + + } else { + + warn "serviceid $serviceid last_srv_date == $currdate; skipping\n" + if $DEBUG; - $prev = $cur; } - warn "done with $serviceid\n" if $DEBUG; - exit; - #end-of-kid + exit; #end-of-kid + } } #foreach my $torrus_srvderive dbh->commit or die dbh->errstr; myexit() if sigterm() || sigint(); + warn "sleeping an hour\n" if $DEBUG > 1; + sleep 60*60; warn "restarting main loop\n" if $DEBUG > 1; - #sleep 60 unless $found; } sub _shouldrun {