httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From George Carrette <George_Carre...@iacnet.com>
Subject Re: memory allocations stuff (question)
Date Wed, 09 Jul 1997 16:25:40 GMT
I'm game to call it ServerRLimitXXX, and to have also ServerRLimitCPUPerRequest
to do the getrusage();setrimit() for each request, although I must warn people
that in the past I've seen Unix versions which get "tired" of having too many 
calls
to setrlimit. Although, I suppose that it was max requests per child is all 
about!

>t's too bad that there's no way to catch a signal when the soft limits
>are reached and respond with a 5XX error.

There is a signal for the soft cpu limit! It works fine, once.
No signal for the memory limit, even soft. 
But do you mean that there is no existing structure in the httpd for
longjumping out of bad situations and responding with 5XX?

One thing about the ServerRLimitCPU and ServerRLimitMEM, since these
must be set sometimes outside of the context of a request, and sometimes
inside the context, where should I grab the values?

Note that I'll have to back burner any work on Apache for a while,
and I think that in practice anything more complex than the limit
code I already pached up gets into serious "diminishing returns."

Now what would be seriously useful is to have the mother of all HTTPD's
record in the error_log when a child has exited with a non-zero value.

That could go a long way toward ensuring software quality.
Right now a developer has to get that info from the pacct file.

Mime
View raw message