Hi,
Has anyone come across problem with directory server having too many fds open, which then causes it to not receive any new connections?
Version: Fedora-Directory/7.1 B2005.146.2010 OS: Linux ds01 2.6.11-1.1369_FC4 #1 Thu Jun 2 22:55:56 EDT 2005 i686 i686 i386 GNU/Linux
From the logs/error:
[28/Jul/2005:19:22:42 +1000] - Listening for new connections again [28/Jul/2005:19:22:42 +1000] - Not listening for new connections - too many fds open [28/Jul/2005:19:22:42 +1000] - Listening for new connections again [28/Jul/2005:19:22:42 +1000] - Not listening for new connections - too many fds open [28/Jul/2005:21:00:22 +1000] - Listening for new connections again
One thing I have noticed is a lot of tcp connections owned by the ns-slapd process, that are suspended in the CLOSE_WAIT state.
# netstat -nap ... tcp 0 0 ::ffff:10.10.246.31:389 ::ffff:10.10.245.12:54566 CLOSE_WAIT 17118/ns-slapd tcp 0 0 ::ffff:10.10.246.31:389 ::ffff:10.10.245.12:54502 CLOSE_WAIT 17118/ns-slapd tcp 0 0 ::ffff:10.10.246.31:389 ::ffff:10.10.245.12:54758 CLOSE_WAIT 17118/ns-slapd tcp 0 0 ::ffff:10.10.246.31:389 ::ffff:10.10.245.12:54569 CLOSE_WAIT 17118/ns-slapd tcp 0 0 ::ffff:10.10.246.31:389 ::ffff:10.10.245.12:54312 CLOSE_WAIT 17118/ns-slapd tcp 0 0 ::ffff:10.10.246.31:389 ::ffff:10.10.245.12:54440 CLOSE_WAIT 17118/ns-slapd tcp 0 0 ::ffff:10.10.246.31:389 ::ffff:10.10.245.12:54379 CLOSE_WAIT 17118/ns-slapd tcp 0 0 ::ffff:10.10.246.31:389 ::ffff:10.10.245.12:54315 CLOSE_WAIT 17118/ns-slapd
Regards,
When I installed FDS the setup script was pretty specific about the need to ncrease the max number of open file descriptors.
This is what I setup:
fs.file-max 64000 in /etc/sysctl.conf
* - nofile 8192 in /etc/security/limits.conf
and in my init.d script: ulimit -n 8192
On 7/27/05, Craig Ayliffe cayliffe@gmail.com wrote:
Hi,
Has anyone come across problem with directory server having too many fds open, which then causes it to not receive any new connections?
Version: Fedora-Directory/7.1 B2005.146.2010 OS: Linux ds01 2.6.11-1.1369_FC4 #1 Thu Jun 2 22:55:56 EDT 2005 i686 i686 i386 GNU/Linux
From the logs/error:
[28/Jul/2005:19:22:42 +1000] - Listening for new connections again [28/Jul/2005:19:22:42 +1000] - Not listening for new connections - too many fds open [28/Jul/2005:19:22:42 +1000] - Listening for new connections again [28/Jul/2005:19:22:42 +1000] - Not listening for new connections - too many fds open [28/Jul/2005:21:00:22 +1000] - Listening for new connections again
One thing I have noticed is a lot of tcp connections owned by the ns-slapd process, that are suspended in the CLOSE_WAIT state.
# netstat -nap ... tcp 0 0 ::ffff:10.10.246.31:389 http://10.10.246.31:389 ::ffff:10.10.245.12:54566 http://10.10.245.12:54566 CLOSE_WAIT 17118/ns-slapd tcp 0 0 ::ffff:10.10.246.31:389 http://10.10.246.31:389 ::ffff:10.10.245.12:54502 http://10.10.245.12:54502 CLOSE_WAIT 17118/ns-slapd tcp 0 0 ::ffff:10.10.246.31:389 http://10.10.246.31:389 ::ffff:10.10.245.12:54758 http://10.10.245.12:54758 CLOSE_WAIT 17118/ns-slapd tcp 0 0 ::ffff:10.10.246.31:389 http://10.10.246.31:389 ::ffff:10.10.245.12:54569 http://10.10.245.12:54569 CLOSE_WAIT 17118/ns-slapd tcp 0 0 ::ffff:10.10.246.31:389 http://10.10.246.31:389 ::ffff:10.10.245.12:54312 http://10.10.245.12:54312 CLOSE_WAIT 17118/ns-slapd tcp 0 0 ::ffff:10.10.246.31:389 http://10.10.246.31:389 ::ffff:10.10.245.12:54440 http://10.10.245.12:54440 CLOSE_WAIT 17118/ns-slapd tcp 0 0 ::ffff:10.10.246.31:389 http://10.10.246.31:389 ::ffff:10.10.245.12:54379 http://10.10.245.12:54379 CLOSE_WAIT 17118/ns-slapd tcp 0 0 ::ffff:10.10.246.31:389 http://10.10.246.31:389 ::ffff:10.10.245.12:54315 http://10.10.245.12:54315 CLOSE_WAIT 17118/ns-slapd
Regards,
-- Craig Ayliffe
-- Fedora-directory-users mailing list Fedora-directory-users@redhat.com https://www.redhat.com/mailman/listinfo/fedora-directory-users
Has anyone come across problem with directory server having too many fds open, which then causes it to not receive any new connections?
This typically indicates that you have a client application that is leaking connections.
The sockets on CLOSE_WAIT shouldn't be a problem. It's active open connections that bung up your server.
If you really do need gazillions of active connections, then you'll need to increase the file descriptor limit as detailed in the documentation.
389-users@lists.fedoraproject.org