I am not an xinetd expert, but have you checked out the cps config parameter to xinetd? By default, xinetd limits the number of incoming connections to 50 per second and if it hits this limit, then it will refuse any connection for 10 seconds. Now, I can't imagine getting 50 connections per second unless you have a large population that is using a web based IMAP client (such as SquirrelMail) that initiates a new connection for every action.
The reason I ask is that the piece of log file you presented indicates that the client disconnected from binc, and not the other way around. Everything looks normal.
What happens after all instances going down? Does it eventually come back up by itself or do you have to do something to restrat binc? Could there be some resource exhaustion? Such as quotas being exceeded or CPU spiking? Would you be able to post your log files on the a Web site and send us the URL the next time this happens?
Regards, Henry
On Sat, 11 Oct 2003 07:47:54 +0200, NOC (EUROFMC) <[EMAIL PROTECTED]> wrote:
Hi everybody,
...slight disturbances with bincimap when a great deal of users are requesting the server : some instances of imap hanghup and, after a few time, all the instances are down.
See the log below :
Oct 3 15:41:49 balance bincimapd[6254]: User <franklin.lecointre> entered authenticated mode.
Oct 3 15:41:50 balance bincimapd[6254]: Main server shutting down - bodies:0 statements:4
Oct 3 15:41:50 balance bincimap-up[6253]: Client disconnected
Oct 3 15:41:50 balance bincimap-up[6253]: Unprivileged stub shutting down - read:227 bytes, wrote:442 bytes.
I'm using bincimap 1.2.2, xinetd 2.3.12 and Redhat 7.3 pro on a e-server IBM.
Any idea ?
Thanks everybody !
Fran�ois
-- Henry Baragar Principal, Technical Architecture 416-453-5626 Instantiated Software Inc. http://www.instantiated.ca
