On Thu, Apr 19, 2001 at 05:31:52PM -0700, Brett wrote:
> Setting up /var/qmail1, /var/qmail2, etc. each with a concurrency of, say,
> 100, as opposed to one install of qmail with the big concurrency remote
> patch with concurrency set to 500.
This might, IMHO, have some advantages, especially if /var/qmail1 and
/var/qmail2 are different spindles. However, see below.
>
> Apparently you have to adjust the linux kernel to get your concurrency up to
> 500 so wouldn't it be easier to just have multiple qmail installs? Is there
> an overall limit of concurrent connections that's unrelated to whether
> you're running one or several qmail installs? Thanks.
Yes, the kernel is the limit -- maximum running processes, maximum open
file descriptors, etc. The kernel limits that you'd likely run into
would be the system-wide ones, not just per-process ones. I don't believe
that kernel recompiles are required anymore, though -- check
/proc/sys/fs/file-max and /proc/sys/fs/inode-max tunables, and 'man
bash' for ulimit options -- kernel options for open file descriptors do
not seem to be hard-coded anymore (since 2.2.12 IIRC).
>
>
--
Greg White
Those who make peaceful revolution impossible will make violent
revolution inevitable.
-- John F. Kennedy
