sub InitMailWatchLogging {
my $pid = fork();
if ($pid) {
# MailScanner child process
waitpid $pid, 0;
MailScanner::Log::InfoLog("Started SQL Logging child");
} else {
# New process
# Detach from parent, make connections, and listen for requests
POSIX::setsid();
if (!fork()) {
$SIG{HUP} = $SIG{INT} = $SIG{PIPE} = $SIG{TERM} = $SIG{ALRM} = \&ExitLogging;
alarm $timeout;
$0 = "MailWatch SQL";
InitConnection();
ListenForMessages();
}
exit;
}
}
sub InitConnection {
# Set up TCP/IP socket. We will start one server per MailScanner
# child, but only one child will actually be able to get the socket.
# The rest will die silently. When one of the MailScanner children
# tries to log a message and fails to connect, it will start a new
# server.
socket(SERVER, PF_INET, SOCK_STREAM, getprotobyname("tcp"));
setsockopt(SERVER, SOL_SOCKET, SO_REUSEADDR, 1);
my $addr = sockaddr_in($server_port, $loop);
bind(SERVER, $addr) or exit;
listen(SERVER, SOMAXCONN) or exit;
# Our reason for existence - the persistent connection to the database
$dbh = DBI->connect("DBI:mysql:database=$db_name;host=$db_host", $db_user, $db_pass, {PrintError => 0});
if (!$dbh) {
MailScanner::Log::WarnLog("Unable to initialise database connection: %s", $DBI::errstr);
}
$sth = $dbh->prepare() or
MailScanner::Log::WarnLog($DBI::errstr);
}
sub ListenForMessages {
my $message;
# Wait for messages
while (my $cli = accept(CLIENT, SERVER)) {
my($port, $packed_ip) = sockaddr_in($cli);
my $dotted_quad = inet_ntoa($packed_ip);
# reset emergency timeout - if we haven"t heard anything in $timeout
# seconds, there is probably something wrong, so we should clean up
# and let another process try.
alarm $timeout;
# Make sure we"re only receiving local connections
if ($dotted_quad ne "127.0.0.1") {
close CLIENT;
next;
}
my @in;
while (<CLIENT>) {
# End of normal logging message
last if /^END$/;
# MailScanner child telling us to shut down
ExitLogging if /^EXIT$/;
chop;
push @in, $_;
}
my $data = join "", @in;
my $tmp = unpack("u", $data);
$message = thaw $tmp;
next unless defined $$message{id};
# Check to make sure DB connection is still valid
#InitConnection unless $dbh->ping;
# Log message
$sth->execute()
if (!$sth) {
MailScanner::Log::WarnLog("$$message{id}: MailWatch SQL Cannot insert row: %s", $sth->errstr);
} else {
MailScanner::Log::InfoLog("$$message{id}: Logged to MailWatch SQL");
}
# Unset
$message = undef;
}
}
I omitted most of the $sth->execute statement because it's just a bunch of variables. Whenever MailScanner scans a message, it calls the function MailWatchLogging. Most of that function is just manipulating variables, but it also tries to write to the server that was started in the code above with this:
while (1) {
socket(TO_SERVER, PF_INET, SOCK_STREAM, getprotobyname("tcp"));
my $addr = sockaddr_in($server_port, $loop);
connect(TO_SERVER, $addr) and last;
# Failed to connect - kick off new child, wait, and try again
InitMailWatchLogging();
sleep 5;
}
Now, the problem I'm seeing is that the fork'd part of the program keeps starting over and over until all the resources on the server are exhausted and I have to restart it. No error is reported in the syslog except for the message "Started SQL Logging child", which you can see is called in InitMailWatchLogging. The number of processes is much greater than the number of emails being scanned, so I believe that the entire problem exists in InitMailWatchLogging, but as previously stated, I don't know how to trace a fork. The first error I mentioned is the only one that I've been able to see, so I had hoped that it had something to do with the original problem. In reply to Re^2: Problems with DBI::trace_msg
by dirtdart
in thread Problems with DBI::trace_msg
by dirtdart
| For: | Use: | ||
| & | & | ||
| < | < | ||
| > | > | ||
| [ | [ | ||
| ] | ] |