Long Lived imapd processes

Lawrence Greenfield leg+ at andrew.cmu.edu
Wed Apr 16 13:00:13 EDT 2003

As I said, you can SIGTERM imapd processes _as long as those imapd
process are connected to a client_. If they aren't connected to a
client, _it won't work_.

In the future, imapds will catch SIGTERM and exit gracefully.

In the slightly further future, we'll apply some variation of a
process accounting patch.


   Date: Wed, 16 Apr 2003 11:01:37 -0400
   From: Earl R Shannon <Earl_Shannon at ncsu.edu>

   I tried the SIGTERM kill method and I am concerned about a behavior I'm 
   seeing. If the last imapd process is removed from the server the master 
   process does not start a new one. The master will accept a connection 
   but it cannot give it to an imapd process. The master thinks at least 
   one must still be around somewhere.

   OK. Doctor it hurts when I do this. Doctor says don't do that. Fine.

   But, seems to me if it could happen then the master should be prepared 
   to handle the situation and it isn't. Should I consider this a bug?
   Things happen and it sounds like at least one person on the list is 
   sending kill -SIGTERM ( kill -TERM for us Solaris users ) to their imapd 
   processes. This seems to be leaving the master with bad info based on 
   the behavior I'm seeing.

   Earl Shannon

More information about the Info-cyrus mailing list