20 messages in net.sourceforge.lists.courier-usersRE: [courier-users] authldap failing ...
FromSent OnAttachments
Bowie BaileyJan 24, 2006 11:36 am 
Dirk KulmseeJan 24, 2006 1:02 pm 
Sam VarshavchikJan 24, 2006 3:57 pm 
Sam VarshavchikJan 24, 2006 3:58 pm 
Sebastian LechteJan 25, 2006 3:22 am 
Alessandro VeselyJan 25, 2006 5:07 am 
Bowie BaileyJan 25, 2006 10:46 am 
Sam VarshavchikJan 25, 2006 3:24 pm 
Sam VarshavchikJan 25, 2006 3:30 pm 
Bowie BaileyJan 26, 2006 5:53 am 
Sam VarshavchikJan 26, 2006 3:02 pm 
Bowie BaileyJan 27, 2006 6:17 am 
Georg LutzJan 27, 2006 2:56 pm 
Sam VarshavchikJan 27, 2006 3:27 pm 
Sam VarshavchikJan 27, 2006 3:32 pm 
Bowie BaileyJan 30, 2006 10:50 am 
Sam VarshavchikJan 30, 2006 3:29 pm 
Bowie BaileyJan 31, 2006 6:10 am 
ecu...@encontacto.netMar 16, 2006 3:20 am 
Georg LutzMar 16, 2006 12:08 pm 
Actions with this message:
Paste this link in email or IM:
Paste this link in email or IM:
Atom feed for this thread
Paste this URL into your reader:
Subject:RE: [courier-users] authldap failing randomlyActions...
From:Bowie Bailey (Bowi@BUC.com)
Date:Jan 30, 2006 10:50:08 am
List:net.sourceforge.lists.courier-users

Sam Varshavchik wrote:

Bowie Bailey writes:

Can you give me a bit of insight into how the authdaemon processes are handling the LDAP connections? Do they ever close them or do they assume the connection will remain open indefinitely? What I would assume is this:

- Each process opens a connection and holds it open. - When an authentication request comes in, it tries to use the connection. - If it fails, the process generates a 450 error and then opens a new connection for the next request.

This would mean that whenever an authdaemon process is idle for more than the LDAP timeout period, I should expect to get an error on the next incoming email that attempts to use that process. Is that accurate?

That's pretty much it. When no authentication requests arrive for five minutes, all open LDAP connections are closed. For this to happen, though, absolutely nothing needs to happen for five minutes.

The thing is that even if requests trickle in every once in a while, each request will go to only one daemon at random; the others will continue to remain idle and hold open connections.

I'm curious if the following patch will work, try it with the timeout set back to 30 seconds:

That worked! After installing the patch, I cannot recreate the failures.

Can you include this with the next release, or will I need to continue patching each upgrade?