Network Card Waking From The Dead

Hi All,

Bit of a weird one this - and not sure if it's even applicable to this forum/site :laugh:

I've got two nix webservers running Apache, MySQL and Heartbeat (plus the rest!). They use MySQL replication and rsync the pages between master and slave and use heartbeat as a failover.

Now when I take the master down, everything is quite happy carrying on from the slave.

But, something wakes the master up automatically - and I'm not sure what it is! It's not a a problem as the way I have heartbeat set-up, the old master stays as a slave until I'm ready to run my auto-takeback scripts - but I would like to sort it!

I think it's a setting somewhere that tells the server to wake up if the network card is pinged (I know you can do this for a windows PC) - but I'm not sure how to go about finding it! I'm runnig SuseLinux 8.2 and the servers are a couple of IBM xSeries 335's. They are connected together by peer-to-peer NIC and serial ports, and both connect to the DMZ of a firewall via NIC (the firwall is behind the router).

If anyone has any ideas what-so-ever - please post!

Why dont you set it up so that they exclusively go through the DMZ, and disconnect the peer to peer stuff. That way you can set it up to disable ICMP packets in your firewall rules. This would definitely keep the ping packets from reaching to your servers from anywhere.

Yeah - this is how I original set it up. But the HeartBeat HA system requires a couple of peer-to-peers (in this case a NIC and serial) for the beat. I need some sort of setting or tick that tells the server not to boot back up if the NIC is pinged but I can't find it!

What is also weird is that the boot process is strange when the server is brought back up by a ping... The loading screen talks about DHCP and seems to wait for a connection before it times out and continues to boot up normally...

Sounds like it is pulling a separate configuration for the box when it's "wake on lan" options are triggered. This sort of makes sense because there is probably no static ip configured for this, therefore it looks for a DHCP server by default. You might want to check out your lilo.conf file, and see whats going on there. See how many boot options are configured for the linux loader, and see if you can spot a wake on lan configuration. I hope this helps.

I am guessing, but during system maintenance it is doing a network reset

Thank you - but it seems that the server is using some sort of grub type critter :laugh: - not lilo. Does anyone have any idea what I am looking for? The grub.conf looks like this:

root (hd0,1)
install --stage2=/boot/grub/stage2 /boot/grub/stage1 d (hd0) /boot/grub/stage2 0x8000 (hd0,1)/boot/grub/menu.lst
quit

but I can't see either the stage2 or stage1 files (they are not text and I don't know what to open them with for editting)
while the grub menu (which I can't see being the cause) looks like this:

color white/blue black/light-gray
default 0
timeout 8

title linux
kernel (hd0,1)/boot/vmlinuz root=/dev/sda2 vga=0x31a splash=silent showopts
initrd (hd0,1)/boot/initrd

title floppy
root (fd0)
chainloader +1

title failsafe
kernel (hd0,1)/boot/vmlinuz.shipped root=/dev/sda2 showopts ide=nodma apm=off acpi=off vga=normal nosmp noapic maxcpus=0 3
initrd (hd0,1)/boot/initrd.shipped

Now I have no idea what any of that means - but is it possible that some sort of wake on LAN setting or network reset could be found here? Or am I looking in the wrong place?

Unless I'm misunderstanding you, the computer turning itself on is a system setting which is OS independent. Check your bios for a 'wake on lan' setting(or similar) and disable it.
If I am misunderstanding, ignore me.

Yeah - I thought that as well - but I checked through the bios settings before posting my problem and couldn't find anything - I'm using a IBM xSeries 335 if that means anything to anyone... :laugh: