Re: Tux serving graphics only

Chris Davies <[email protected]>
Newsgroups gmane.network.tux
Message-ID <[email protected]>
On Tue, 2003-08-26 at 10:10, Marco Rothley wrote:
> you might have problems with logfiles (if you need them). You will have two 
> different logfiles and - even worth - there is still a bug in TUX that causes 
> the whole machine to hang about once a week (on some installations). As far 
> as I know there isn't any patch to solve this problem, yet, and nobody knows 
> where the bug is.

we have two servers that serve over 20 million pageviews with a ton of
graphics every month.  Uptime on each is 70+ days which is when the
machines were put in service.

All html is passed to an apache2 backend, parsed, and sent to the
surfer.

Each machine serves roughly 10000gb/month and no lockups yet.

We don't run logging -- which has been suspected of causing the problem,
however, we have run logging a few times for multiple days to verify
things were working as expected.

It just works.  Context switches when the machine is pushing 35mb/sec
are in the 1100/second range compared with the 4800/second when we were
running only apache.

I really haven't found tux to be that much faster on a lightly loaded
server, but it scales wonderfully.  At 35mb/sec, we're using 8% of the
CPU.  Neither machine is especially powerful as they were supposed to be
proof of concept machines and the client was impressed enough that we
never moved off the demo boxes.  One is a P3/800mhz with 512mb ram and a
36gb 10kRPM scsi drive and the other is a dual 1ghz with 1gb ram.

A kernel panic is relativly easy to diagnose if someone that has the
problem can reproduce the problem.  Even if it cannot be triggered on
demand, enough logging can be put into place to track it down reasonably
quickly.  Comparison of the options that they use with ones that don't
hang would be a good first step.

/etc/sysctl.tux
net/tux/virtual_server = 3
net/tux/clientport = 81
net/tux/referer_logging = 0
net/tux/logging = 0
net/tux/max_keepalives = 1000
net/tux/keepalive_timeout = 30
net/tux/nonagle = 0
net/tux/TDprintk = 0
net/tux/Dprintk = 0
# these do seem to make it faster
net/tux/generate_cache_control = 1
net/tux/generate_etags = 0
net/tux/generate_last_mod = 0
net/tux/defer_accept = 1

I don't do any CGI on the server, but the above settings are what we
tweaked and both have run flawlessly since day 1 of turning things live.

perhaps we can figure out where the problem is by comparing notes.
lmpx.com only provides a reader for public news (NNTP) servers. It is not affiliated with the servers or forums shown here and is not responsible for the content of articles, which is written by their respective authors.