The program has been running for more than 10 years without this beeing an
issue - at least I have never been aware of this kind of problem.
Perhaps I only first see it now on the new server because the server is a
virtual machine and maybe doesn't have a dedicate cpu and more often will
get interrupts on the cpu.
The reason that the program uses processes is that is has a mother process
fetching work from the database and communicating the work/result of the
work to/from the threads via Thread::Queue.
Am I the first one using threads doing http requests in perl on linux
having this problem? Seems odd.
If I would change my program to using processes instead of threads, would
gethostbyname then be process safe?
It looks as if gethostbyname has a version that is reentrant:
gethostbyname_r.
Any reason that perl doesn't use that version?
Regards
On 29 May 2013 10:12, Francesco Nidito <francesco.nid...@gmail.com> wrote:
> Unfortunately, gethostbyname is not reentrant.
>
> (maybe a stupid question) Is there any particular reason why you cannot
> use multiple processes instead?
>
> Typed on a very small keyboard...
> ------------------------------
> From: Jesper Persson <jesper.j.pers...@gmail.com>
> Sent: 29/05/2013 07:17
> To: libwww <libwww@perl.org>
> Subject: is IO:Socket:INET thread safe?
>
> Hi everybody,
>
> I have a multithreaded program that monitors the content of web pages.
> On a newly installed CentOS server I have seen strange behaviour where I
> get 404 because http request is sent to a wrong ip address not maching the
> domain that was requested.
>
> I have inserted a print statement in IO:Socket:INET.pm in the procedure
> _get_addr
> to see what ip address the domain is translated to:
>
> sub _get_addr {
> my($sock,$addr_str, $multi) = @_;
> my @addr;
> if ($multi && $addr_str !~ /^\d+(?:\.\d+){3}$/) {
> (undef, undef, undef, undef, @addr) = gethostbyname($addr_str);
> } else {
> my $h = inet_aton($addr_str);
> push(@addr, $h) if defined $h;
> }
> open(FH, ">>/tmp/socket_get_addr.log");
> print FH localtime().": addr_str: $addr_str, addr:
> ".inet_ntoa($addr[0])."\n";
> close(FH);
> @addr;
> }
>
> From the log file:
> Wed May 29 07:38:09 2013: addr_str: dk1.siteimprove.com, addr:
> 93.160.60.57
> Wed May 29 07:38:09 2013: addr_str: siteimprove.dk, addr: 93.160.60.57
>
> the ip address of siteimprove.dk is wrong and at the same second just
> before a lookup of dk1.siteimprove.com gets the same ip address and that
> is correct.
>
> 7 minutes later in the log file the siteimprove.dk domain is looked up
> correctly:
> Wed May 29 07:45:11 2013: addr_str: siteimprove.dk, addr: 93.160.60.87
>
> Manually looking up the ip on then server gets the right ip address.
>
> this entry is in the hosts file:
> 93.160.60.87 siteimprove.dk
>
> CentOS release 6.3 (Final)
> Summary of my perl5 (revision 5 version 10 subversion 1) configuration:
> IO:Socket:INET version: $VERSION = "1.31";
>
> Any help would be greatly appreciated.
>
> Regards
> Jesper Persson
>
