[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
WWW
- To: outages-list@eskimo.com
- Subject: WWW
- From: Robert Dinse <nanook@eskimo.com>
- Date: Sat, 22 May 1999 21:28:27 -0700 (PDT)
- Resent-Date: Sat, 22 May 1999 21:28:31 -0700
- Resent-From: outages-list@eskimo.com
- Resent-Message-ID: <"0QBUV1.0.i37.lFuHt"@mx1>
- Resent-Sender: outages-list-request@eskimo.com
Regarding WWW:
Aaron got Apache recompiled to use glibc and things seems to be
working much better now.
Technical discussion follows for those that are interested or might
be able to offer suggestions:
First, the main web server runs on a 4-CPU Sparc SS-10 with quad
RTK-625 CPU's running what was called SparcLinux and then when Sun got
bent out of shape got changed to S-Linux, and then as the Ultra port
became mainstream became UltraLinux even if it's not UltraSparc hardware
you're running it on.
We're still on version 2.0.35 of the kernel, later versions of 2.0.x
were unstable on that platform, 2.1.x has been unstable on that platform,
and so far the 2.2.x kernels we've tried have been unstable but we've only
tried the very early ones.
Until yesterday, we were using libc5 libs. Yesterday Aaron installed
glibc and the compatibility libs. Web server didn't run at all well under
the compatibility libs.
Today he recompiled to use glibc and it seems to be back to working
well. NIS is also working fine as a client. However, if I attempt to run
a ypserver on that machine, and I wanted to do this because it would
decrease network traffic leaving more bandwidth for serving web pages, it
flakes out.
If ypbind is bound to another ypserver, I can do:
ypcat passwd.byname | wc -l
And always get the whole file, every time.
Right now I have this setup as just an NIS client since I can not get
ypserv to function reliably but would like to get that fixed since
eliminating any unnecessary network traffic will leave that much more
bandwidth for serving web pages.
Once we get this straightened out, we'll revisit the 2.2 kernel.
There have been some significant advancements in how resources are handled
in a multi-CPU environment that should, in theory at least, result in more
effecient operation, and I am told the NFS problem is fixed with kernel
NFS. Only time will tell.