httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Manoj Kasichainula <>
Subject Re: MPM: HARD_SERVER_LIMIT=8, scoreboard layout
Date Sat, 10 Jul 1999 06:15:26 GMT
On Sat, Jul 10, 1999 at 04:06:56PM +1000, Brian Havard wrote:
> Is there any good reason for HARD_SERVER_LIMIT being such a small number?
> In the mpm I'm writing, server == thread so it needs to be a big number.

Dean presumably started the MPM code base from the stuff in the
apache-apr tree, where HARD_SERVER_LIMIT refers to processes.

HARD_.*_LIMIT should really be moved into the MPM itself.

> Also, it seems the layout of the scoreboard is going to differ for each mpm.
> How are we going to handle this? It's not a problem as long as the mpm module
> is the only one accessing the scoreboard but once mod_status etc are brought
> in it gets a bit harder.

I think it will be impossible for mod_status to be portable between
MPMs, simply because it has to know so many details about specific
MPMs. So, each MPM would have its own mod_status (and mod_info). How
this is organized will be a fun question, though. Maybe the MPM should
build the correct hooks so that it is its own status module. (Ow! that

> How about keeping the scoreboard.h in the mpm's directory and adding that to
> the include path?

This is one of the things I've been wanting to do. And the MPM's
directory is already in the include path; the pthread MPM stashes
include files there now.

> You could also put mpm specific settings like
> HARD_SERVER_LIMIT in there too so that values sensible for each mpm can be
> set.

The whole concept should be MPM-specfic. One MPM will use process
limits only, one will use process and thread limits, one will use only
thread limits, one may have no limits at all.

Manoj Kasichainula - manojk at io dot com -
"This is not a novel to be tossed aside lightly.  It should be thrown with
great force." -- Dorothy Parker

View raw message