Hi,

> I've got a machine that every time it reboots from a power failure Apache
> gets hung. I've not been able to track down the problem to Apache as the
> Apache logs don't indicate any errors. If I SSH into the box and stop
> apache and then restart it, it works fine. Also, sometimes if I do a
> reboot of the box Apache will start fine (but sometimes it won't).

Ok. What do you consider hung? Are you trying to fetch sites using a
Windows system? If so, you may be misdiagnosing - Windows will time out
even if a TCP/IP handshake is established.

> Everytime the box reboots from a power failure, Apache hangs (or really,
> just doesn't serve pages, it actually appears to start just fine, its
> just deaf).

It sounds like a DNS problem. Is there any chance that Apache gets
launched before BIND (or whatever DNS server) comes up? Or before your
Internet connection is established, if you use remote DNS servers?

When it's hung, what happens if you telnet to port 80 directly? What
happens if you wait a few minutes?

> I did notice that in my dmesg, the last line is "sn0: receive FIFO
> overrun". I know sn0 is my ethernet port. Why would I get a FIFO overrun?
> Would this have anything to do with Apache not starting properly? Is the
> FIFO overrun something I should be worried about, and if so, any ideas
> how to fix it?

That's "normal". The motherboard ethernet only has 16k of buffering, so
sometimes it happens, especially when the kernel is doing time consuming
things like handling pseudo-DMA SCSI transfers while the network is busy
/ noisy.

I use many of my Quadras as IP NAT / IPv6 routers, and prefer to connect
Nubus ethernet cards with 32k or 64k of buffers on the noisier connection,
which is usually the Internet connection. But many autoswitching ports
won't negotiate with Nubus 10-base-T cards, but will with motherboard
ethernet, so sometimes I don't have a choice. No big deal - they're just
warnings.

John Klos
Sixgirls Computing Labs