On Thu, 15 Nov 2001, Sam Varshavchik wrote:

=>Slepp Lukwai writes: => =>> Ignore me if this is a FAQ, but I haven't seen any information about it =>> yet... =>> =>> I have one mailbox with a meagre 2505 messages in it at the moment, and =>> when I issue a server side sort of the mailbox based on thread/subject =>> from PINE, it goes for a bit and imapd uses 100% CPU on the mail server. =>> After about 10 seconds, imapd abruptly closes the connection reporting the =>> sort failed. Any ideas on this one and what might be causing it? I don't =>> think 2500 messages is such a bad thing to request a sort of.. => =>You forgot to specify: => =>A) The server version you are using. => =>B) How much memory you've allocated to the server process.

Oh.. Yes. Forgive me for being entirely inept, it happens when I haven't slept enough. :>

Anyhow... I use Courier-IMAP 1.3.9 formerly, and as of today, I use 1.3.12.

It's running with qmail-1.0.3 with the LDAP patches (hence using authldaemon.ldap).

The pine client connects to the server using SSL (mail/user=slepp/ssl/novalidate-cert) and is version 4.33.

The ulimit set in the configuration files is 65536, which should be a nice 64 megs of RAM.

The mailbox I'm sorting is 1154k in size, with a total of 2519 messages at this point (it's a mailing list).

The sort succeeds on the smaller mailboxes, in the ranges of under around a thousand, but above that point it tends to fail.

If imapd is taking more than 64 megs to sort 2500 messages, then something needs to be looked at.. But I don't think that's the case.

This problem, and a strange one with authorization failure on first attempt guarantees every auth attempt after that fails (but I'm pretty sure I know why), are the only two issues I've had with Courier IMAP in the past two years or so...

The log messages from Courier into the syslog seem to be entirely useless when it comes to this problem I'm having, and since log messages don't seem to be dynamically configurable (except perhaps warn/alert) I can recompile if required for the debugging information.