Re: Fatal error negotiating setup
David Severance <[email protected]> Fri, 03 Jun 2011 18:35:16 -0700
| Newsgroups | gmane.mail.perdition.user |
|---|---|
| Message-ID | <[email protected]> |
On 5/4/2011 12:43 AM, Simon Horman wrote: > Sorry, I've been a bit slow. > I got the email and I'll try and get to it soon. Well I believe I have finally tracked down the problem. I took the thoeries you gave me along with the suggestions the Panda Imap developer gave me and tried to setup some testing. It took awhile but last week I finally caught the problem and found a clue in the messages file. It logged a failure in xinetd. If you recall I mentioned I was working on RHEL5. It was actually a migration project from RHEL4 to RHEL5 and it turns out there is a change in the way xinetd functions that impacts our environment in particular. In the version of xinetd provided in RHEL5 it includes some new (as compared to the RHEL4 version) keywords. One of them is per_source and is limited to 10 connections per source ip a second. For an imap service that gets all it's connections from a proxy like perdition this just won't work. Of course you won't notice until you load it up with at least half a dozen very active users but when you do it will just terminate any additional connections. So I set per_source = unlimited to get the previous behavior. After a week of additional testing with double the number of users all seems well again. Thanks for the insights, they did help me when more clues finally did emerge. thanks, David -- David Severance Central Computing Services Office of Information Technology (949) 824-7552 [email protected] ______________________________________________ Perdition-users mailing list [email protected] http://lists.vergenet.net/listinfo/perdition-users