Re: [Exim] Performance bottleneck scanning large spools.

Quoth Philip Hazel on Tue, Jan 04, 2000:
> I think I understand what you are saying, but I confess to complete
> ignorance about shared memory segments. Obviously I can go away and
> learn about them. Are they "standard" in all the versions of Unix that
> Exim runs on?

I'm not sure about it. I'm not sure all the OSes Exim runs on
support SysV shared memory. Even those that do may not have the
support compiled into the running kernel.

> Fine, a queue runner creates the shared segment, but who updates the
> shared segment as messages arrive and depart, so that it reflects
> reality when the next queue runner process finds it already exists?

No-one, but the queue runner will update it when it wakes up
next. But then, what happens if you run queue by hand or have
more than one queue runner process?

Anyway, if it's only to speed up the process of delivering mail,
consider this: in any case, if the directory is big enough, on
most OSes it will take too much time just to scan the directory
to get to the correct inode.

> My big worry is that messages will get "lost" (not listed in the
> segment) and therefore not delivered because nothing notices them.

Could happen, I believe.

Vadik.

--
The ill-formed Orange
Fails to satisfy the eye:
Segmentation fault.