| Subject: | Connections initiated too fast might result in "Cannot assign requested address" |
Originally reported by Csongor Fagyal at
http://bugs.mysql.com/bug.php?id=26112
Description:
I get periodical errors saying:
Can't connect to MySQL server on '192.168.0.200' (99)
Happens if I try to connect to the server very frequently (in a loop).
Tracing the program gives:
restart_syscall(<... resuming interrupted call ...>) = 0
time(NULL) = 1170758829
socket(PF_INET, SOCK_STREAM, IPPROTO_IP) = 3
fcntl64(3, F_SETFL, O_RDONLY) = 0
fcntl64(3, F_GETFL) = 0x2 (flags O_RDWR)
connect(3, {sa_family=AF_INET, sin_port=htons(3306),
sin_addr=inet_addr("192.168.0.200")}, 16) = -1 EADDRNOTAVAIL (Cannot
assign requested address)
shutdown(3, 2 /* send and receive */) = -1 ENOTCONN (Transport
endpoint is not connected)
close(3) = 0
write(1, "Doesn\'t work. Reconnecting in 1 "..., 51) = 51
time(NULL) = 1170758829
rt_sigprocmask(SIG_BLOCK, [CHLD], [], 8) = 0
rt_sigaction(SIGCHLD, NULL, {SIG_DFL}, 8) = 0
rt_sigprocmask(SIG_SETMASK, [], NULL, 8) = 0
nanosleep({1, 0}, {1, 0}) = 0
time(NULL) = 1170758830
socket(PF_INET, SOCK_STREAM, IPPROTO_IP) = 3
fcntl64(3, F_SETFL, O_RDONLY) = 0
fcntl64(3, F_GETFL) = 0x2 (flags O_RDWR)
connect(3, {sa_family=AF_INET, sin_port=htons(3306),
sin_addr=inet_addr("192.168.0.200")}, 16) = -1 EADDRNOTAVAIL (Cannot
assign requested address)
shutdown(3, 2 /* send and receive */) = -1 ENOTCONN (Transport
endpoint is not connected)
How to repeat:
use DBI;
use strict;
my $dsn = "DBI:mysql:database=test:host=192.168.0.200";
my $dbh;
my $user = 'test';
my $password = 'test';
for (1..50000) {
$dbh = DBI->connect($dsn, $user, $password, {'RaiseError' => 1} );
my $sth = $dbh->prepare('SELECT * FROM users');
$sth= undef;
$dbh->disconnect;
$dbh= undef;
}
Suggested fix:
I am thinking whether I run out of ports because of the high number of
connects... I see a lot of "TIME_WAIT" TCP connections when I do netstat
-n (somewhere around 15000). Also, if at a failure I wait a few seconds,
I am able to connect again.
I have found this (a similar situation?):
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=4294599
Or could it be related to bug # 20629 ?
[6 Feb 2007 11:32] Csongor Fagyal
Actually I am pretty sure now that I just simply run out of available
TCP ports, so I have changed severity to "feature request".
However, I would like to see DBD::mysql to handle this situation in a
more transparent way (some sort of a "retry connect" feature).