Comments on these 2 ideas

[Jim Jagielski <ji...@devsys.jaguNET.com>]

I'm looking at making 2 "small" changes but the first could affect users
who use our default conf files and the 2nd is TAOYOD (The Addition
Of Yet Another Directive).

The first relates to MaxRequestsPerChild. Our conf files have this set
to 30, yet the hard-coded default if no directive exists is 0. Shouldn't
these be consistant? If so, which should we change?

Second, it would be "nice" to adjust SCOREBOARD_MAINTENANCE_INTERVAL
without needing to recompile. Granted, this could be considered fluff,
but I would think this useful especially for pre-compiled binaries.
Should I bother generating a patch?
-- 
===========================================================================
   Jim Jagielski   |||   jim@jaguNET.com   |||   http://www.jaguNET.com/
            "That's no ordinary rabbit... that's the most foul,
            cruel and bad-tempered rodent you ever laid eyes on"

Re: Comments on these 2 ideas

[Harrie Hazewinkel <ha...@jrc.it>]

Dirk-Willem van Gulik wrote:
[snip]
> 
> > Second, it would be "nice" to adjust SCOREBOARD_MAINTENANCE_INTERVAL
> > without needing to recompile. Granted, this could be considered fluff,
> > but I would think this useful especially for pre-compiled binaries.
> > Should I bother generating a patch?
> 
> Harrie,
> 
> This is one for you; Does this not fit in with that SNMP thing
> about increasing log and status levels when you actually want to
> dynamically  monitor more closely when you suspect trouble. ???

I am trying to see if I can set certain values during run-time
via the SNMP-agent. The idea is to shift the variable in mind
to shared memory (scoreboard) which are then set during
configuration by the main process as now is done.
The access rights of the SNMP-agent will then be read-write
and the access-rights of all chidren including the main process is
read-only.

I am trying to make it work for the variable "ap_extended_status".
But I am thinking of more variables such as, 
log levels, MaxRequestsPerChild, MaxServer, KeepAlive, 
KeepAliveTimeout, MaxKeepAliveRequests, 
MinSpareServers, MaxSpareServers, MaxClients. 

Maybe a possibility is temporary stopping and starting
a particular virtual host. A while ago I look already
into this and failed miserable, due to lack of internal
apache understanding I guess.


Harrie
0- Harrie Hazewinkel --------------------------------------0
 mailto:harrie.hazewinkel@jrc.it       phone:+39+0332786322
 http://operations.ceo.org/~harrie/      fax:+39+0332785641
 postal: JRC of the E.C.  -       -  Ispra 21020 (VA) Italy
0----------------------------------------------------------0

Re: Comments on these 2 ideas

[Dirk-Willem van Gulik <di...@jrc.it>]

Jim Jagielski wrote:

> The first relates to MaxRequestsPerChild. Our conf files have this set
> to 30, yet the hard-coded default if no directive exists is 0. Shouldn't
> these be consistant? If so, which should we change?

Which IMHO is 'low' these days. I would prefer both to be in the order
of a couple of 100, say 500. Even with SunOS which leaks like a sieve
that is still an order of magnitude away from trouble area's. Or are
there other, worse, platforms ?
 
> Second, it would be "nice" to adjust SCOREBOARD_MAINTENANCE_INTERVAL
> without needing to recompile. Granted, this could be considered fluff,
> but I would think this useful especially for pre-compiled binaries.
> Should I bother generating a patch?

Harrie, 

This is one for you; Does this not fit in with that SNMP thing 
about increasing log and status levels when you actually want to 
dynamically  monitor more closely when you suspect trouble. ???

Dw.

Re: Comments on these 2 ideas

[Dean Gaudet <dg...@arctic.org>]

On Tue, 20 Apr 1999, Jim Jagielski wrote:

> I'm looking at making 2 "small" changes but the first could affect users
> who use our default conf files and the 2nd is TAOYOD (The Addition
> Of Yet Another Directive).
> 
> The first relates to MaxRequestsPerChild. Our conf files have this set
> to 30, yet the hard-coded default if no directive exists is 0. Shouldn't
> these be consistant? If so, which should we change?

Make them both 0.  Screw the ancient platforms which have memory leaks --
put a comment in there for those.

Dean