Martin Husemann wrote:
> On Fri, Jul 29, 2005 at 11:42:04AM +0200, Joel CARNAT wrote:
> 
>>You have two options:
>>
>>    /etc/postfix/main.cf:
>>        smtp_connection_cache_on_demand = no
>>
>>or use an operating system that does not break under stress. 
> 
> 
> Do you also get this log entry?
> 
> postfix/scache[21161]: fatal: event_loop: select: Bad file descriptor
> 

Nope, none such :(

Looks like they go coupled :
postfix/scache[25761]: warning: scache_save_endp_service: unable to 
receive file descriptor
postfix/smtp[6325]: warning: problem talking to service private/scache 
socket: Invalid argument

There are thousand like those and they always come by 2.

> The scache process then exits and the server side can't talk to it any more.

 From what I saw (in the urge of recovering the service ;), the mails 
were still accepted (from smtpd I guess) and began to be processed. But, 
none were delivered, nor appeared in `mailq`. They were only visible 
from "/var/spool/postfix/active/". Then, when I set 
"smtp_connection_cache_on_demand = no", all those mails began to get 
delivered.

> 
> I also got (earlier) a few
> 
> postfix/scache[21161]: warning: scache_multi_drop_endp: close(0): Bad file descriptor
> 
> Looks more like a bug in postfix to me... [still investigating]
> This is probably worth a pkg PR.
> 
> Martin