Stas Bekman wrote:
I think this misleading error is really a bug in Apache:
[Mon Apr 26 15:28:44 2004] [error] server reached MaxClients setting, consider raising the MaxClients setting
It sounds like a one-off bug to me.
the only issue I see is that now both 1.0 and 2.0 environments two servers by default instead of one, which seems noteworthy I guess. I can't see the harm, but when I changed from httpd -X to StartServers 1 I did so with trepidation. at the very least I think we should support a way to get back to a single server if users desire it. -minclients=0 (so maxclients = minclients + 1 = 1) not only isn't intuitive, but it doesn't look to be supported in the code. maybe a -singleserver option is worth adding, which would effectively restore everything you propose to undo here?
That's a different issue, this patch just changes more than one thing. I don't think that setting maxclients = minclients + 1, makes it any different from before. It's the SpareMin/Max and StartServers that may need to be tweaking to preserve the previous state.
I've reviewed again the config change, and the only difference from before are the prefork and apache1.3 settings, and they are the same, so before we had:
<IfModule prefork.c>
StartServers 1
MaxClients @MaxClients@
MaxRequestsPerChild 0
</IfModule>There was one server starting and more added when needed up to MaxClients. So if you wanted to have only one client, you'd say: -maxclients 1
The new change is exactly the same:
<IfModule prefork.c>
StartServers @MinClients@
MinSpareServers @MinClients@
MaxSpareServers @MinClients@
MaxClients @MaxClients@
MaxRequestsPerChild 0
</IfModule>if you say -minclients 1, then you will ever have 1 server, as min/max are set to 1 too. Unless your testing setup requires 2 servers, in which case two servers will start because MaxClient == 2. But this is OK, since you won't be able to run that test with 1 server (e.g. proxy test), as it'll just hang. Keeping min/maxSpare == startservers, prevents from Apache to spawn more than StartServers servers.
Any flows in my logic?
maybe just calling t_server_log_is_expected() early on is a better idea (if it's even possible)?
You mean to workaround the Apache bug? I don't think it'll work. You may end up covering real errors as 'ok', since that Apache error message is not under your control.
Actually this change has a positive side effect, if before setting to -maxclients=1, you'd have tests requiring 2 servers (e.g. proxy) will make your whole suit hang, now it'll work just fine.
finally 'maxclients = minclients + 1 ' is only a default, you have a complete flexibility to set maxclients explicitly, so if you really want to, nothing stops you from saying -maxclients=1 -minclients=1.
__________________________________________________________________ Stas Bekman JAm_pH ------> Just Another mod_perl Hacker http://stason.org/ mod_perl Guide ---> http://perl.apache.org mailto:[EMAIL PROTECTED] http://use.perl.org http://apacheweek.com http://modperlbook.org http://apache.org http://ticketmaster.com
