Also, I have this :
# Controls whether core dumps will append the PID to the core filename.
# Useful for debugging multi-threaded applications.
kernel.core_uses_pid = 1
net.ipv4.ip_local_port_range = 1024 65000
fs.file-max = 128000
net.ipv4.tcp_keepalive_time = 100
as my sysctl.conf. Does this contribute to the problem?
On Sat, Mar 1, 2008 at 2:32 AM, Rich Megginson <rmeggins(a)redhat.com> wrote:
M Vallapan wrote:
> Thanks ! the settings you mentioned work, but only for some time then
> the problem arises again. then I have to manually restart fedora-ds to
> break off all the idle sessions for it to be okay again for a little
> while. How do I go about this ?
>
First, figure out what the clients are which are grabbing all of the
available connections and not letting them go . . .
The server does not close idle connections until some other connection
is made. So you could use ldapsearch to write a script that "pings" the
server every few minutes to force it to close idle connections.
>
> On Wed, Feb 27, 2008 at 1:31 AM, Rich Megginson <rmeggins(a)redhat.com> wrote:
>
>> Low Kian Seong wrote:
>> > Wow ... a bit of ip information there could someone please take out
>> > the last email i sent ? How do i request an email be removed ?
>> >
>> And in your reply, you copied the entire previous message - I've
>> contacted Red Hat support to remove the messages from the archive. But
>> there is no way to revoke the messages once they are sent.
>>
>> This information is interesting:
>>
>>
>> ----- Total Connection Codes -----
>>
>> B1 11480 Bad Ber Tag Encountered
>> U1 5877 Cleanly Closed Connections
>> T1 2187 Idle Timeout Exceeded
>>
>> B1 usually means the client just exit()'ed without first calling close()
>> or shutdown() on the TCP/IP socket. Which is fine. It's the T1 which
>> are odd. Of these 2187, 1864 come from the same client:
>>
>> 13800 XXX.XXX.XXX.129
>>
>> 8254 - B1 Bad Ber Tag Encountered
>> 3608 - U1 Cleanly Closed Connections
>> 1864 - T1 Idle Timeout Exceeded
>>
>> Take a look at the access log where you get the T1 error upon
>> disconnect. You want to find out what the conn=XXXXX is. From there,
>> go back in the access log looking for the operations on that
>> connection. What are they? What application are they from? Why is
>> that application opening connections and just leaving them open? If it
>> is a monitoring application like nagios, you will need to increase the
>> idle timeout for that application. You can do this by using a dedicated
>> BIND dn for that application, then you can increase the idle timeout for
>> that user without affecting any of the other users - see
>>
http://tinyurl.com/2sy8bl
>>
>> If you have a lot of applications that open connections and leave them
>> open for a long time, you will need to figure out how many file
>> descriptors you need for other clients, and you will need to increase
>> the number of file descriptors available for the directory server as
>> well as the size of the directory server connection table -
>>
http://tinyurl.com/35qddb and
>>
http://directory.fedoraproject.org/wiki/Performance_Tuning#Linux
>>
>> See
http://tinyurl.com/35qddb for real time server connection monitoring
>> information.
>>
>>
>>
>> --
>> Fedora-directory-users mailing list
>> Fedora-directory-users(a)redhat.com
>>
https://www.redhat.com/mailman/listinfo/fedora-directory-users
>>
>>
>>
>
> --
> Fedora-directory-users mailing list
> Fedora-directory-users(a)redhat.com
>
https://www.redhat.com/mailman/listinfo/fedora-directory-users
>
--
Fedora-directory-users mailing list
Fedora-directory-users(a)redhat.com
https://www.redhat.com/mailman/listinfo/fedora-directory-users