> On Thu, 9 Nov 2006, Joel Reicher wrote:
> > Is there any data on the extent to which imapd needs to be load balanced?
> > Is it just because of all the fork()ing?
>
> imapd needs little CPU, but intensive I/O and enough memory to prevent
> swapping.
>
> I agree than fork()ing is a problem. fork()ing can be abolished by:
> (1) using Linux or BSD (not using SVR4 or OSF/1)
> AND
> (2) not requiring the use of mlock, one of:
> a) deliver mail to a user-owned directory (e.g., the home directory)
> and not use the spool directory at all
> OR
> b) protect the spool directory 1777 so .lock files can be created
> directly without using mlock.
Interesting. My IMAP server is a NetBSD box, and because I don't have
many users I use a 1777 spool directory. Moreover, all my users have
mbx INBOXes in their homedirs so imapd snarfs.
Nevertheless, I still see what looks like one persistent imapd process
per user (along with others appearing and disappearing when the user does
particular operations).
So when you say "fork()ing can be abolished" do you mean per user or
in general? Should I not be seeing these persistent imapds?
Cheers,
- Joel
_______________________________________________
Imap-uw mailing list
[email protected]
https://mailman1.u.washington.edu/mailman/listinfo/imap-uw
