httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From r..@ai.mit.edu (Robert S. Thau)
Subject Re: linux - apache - another patch
Date Thu, 31 Aug 1995 12:18:49 GMT
Aram --- this code looks a bit weird, and it isn't obvious to me what
it is supposed to accomplish.  Is it supposed to get the process scheduled
every once in a while, by making sure that something runs at least once
a second?  If so, then the problem is scheduler weirdness, but I thought
that had been ruled out as a cause of the Linux problems.

If, on the other hand, the root cause of the problem is the SOMAXCONN
thing (accept() queue jammed with partially negotiated connections; must
wait for them to time out) --- something which has given rise to the
same symptoms on numerous other symptoms --- then this code won't help.

(I actually tried something like this when I was trying to figure out
what the heck was going on with SOMAXCONN, on the theory that there was
some sort of race condition in the kernel which would be broken if 
accept() were periodically reentered.  It didn't work, since that wasn't
the problem...).

rst

Mime
View raw message