pls help: IP migration via ssh: 2 eths & iLO - apache server

stardotstar

hi guys,

I am needing to do a remote controlled ip migration on my server.

Basically I have been running as three nodes on a class c network from my co-lo host

I have two eth NICs - eth0 and eth1 as well as my iLO interface

At the moment I can ssh in on either nic

What has been suggested is that I bind the new IPs to the existing nics and then allow dns propagation to occur before removing the old addresses (which will become inactive at midnight in two days anyway)

Here is my theory (and what i have been trying)

1) ssh in on eth0 legacy addy
2) confirm that I can ssh in on eth1 legacy addy
3) log off from eth1 ssh session
4) edit /etc/ssh/sshd_config to "listen" on the new IP for eth1
5) edit /etc/conf.d/net to bind the new addy,netmask,brd to eth1
6) restart eth1
7) confirm eth1 pings on legacy and new addys

restart sshd
9) confirm ssh still works on eth0 and eth1 legacy (as well as having iLO open as a fallback)
10) ssh in on the new addy on eth1

these 10 things I have tried to complete but only manage to get as far as pinging the new addy on eth1.
I followed this guide:
http://www.gentoo-wiki.info/HOWTO_IP_Aliasing

and can now locally ping the interface eth1 on the newly assigned IP but when I ifconfig I don't get the eth1:1 as shown in the guide.

I am pretty sure that exposing my IPs is the same as running an apache server anyway - so I am presenting some of the results here to get help:

Exil · Posted: Mon Jul 06, 2009 7:34 am Post subject:

emerge iproute2 and then

ip a s

it will show you all ip addresses assigned to interfaces

stardotstar · Posted: Mon Jul 06, 2009 7:45 am Post subject:

OK, thanks for the instruction; I have emerged this tool and provide the following results:

Sysa · Apprentice Joined: 16 Mar 2005 Posts: 161 Location: Europe

stardotstar · Posted: Mon Jul 06, 2009 8:54 am Post subject:

Thank you Sysa.

This is most helpful.

I am finding still some problems:

Sysa · Apprentice Joined: 16 Mar 2005 Posts: 161 Location: Europe

First of all I have to understand your network topology. Please fix me if I'm wrong.
So, I suggest that current (old) connection is using eth1 (202.130.34.115/24 gw 202.130.34.113) and you plan to migrate to the 119.63.202.[186,187]/24 assigned to different NICs (gw 119.63.202.1).
Also it looks like both your NICs are connected to the same network segment and both (old and new) IP addresses and gateways are available now. Please double check that your firewall settings are correct or it is switched off.

stardotstar · Posted: Tue Jul 07, 2009 2:02 am Post subject:

Thank you sysa for your guidance. It is very very much appreciated.

Due to the time variance we did not manage to stay in sync on this in a very economical way and I have managed to progress to a much more advanced but just as borked state as I could ever have hoped! :lol:

This has led to an extended outage that appears to be a DNS issue now.

In case you are able to further assist (and I appreciate the time you have taken so far to "learn" me

)

The topology is this;

Original/Legacy/Old state:

iLO 119.63.202.190 snm 255.255.255.0 gw 119.63.202.1 bc 119.63.202.255
eth0 119.63.202.186 snm 255.255.255.0 gw 119.63.202.1 bc 119.63.202.255
eth1 119.63.202.187 snm 255.255.255.0 gw 119.63.202.1 bc 119.63.202.255

I am learning why to only use one nic - this is obscure to me - my original setup was on the basis that I have only one physical server with two NICs and I therefore ran
ns1.sourcepoint.com.au 119.63.202.186
ns2.sourcepoint.com.au 119.63.202.187

the only use for the second NIC in my thinking was to have a separate physical IP for the secondary or slave name server (I know they are supposed to be on different boxes, let alone separate networks etc... I have a second server waiting to be installed and put in another remote colocation and once I do that I will have a more ideal situation - for now this is what I have...)

Now the goal state is to have all the virtual hosts on the server running as before (above) but like this:

iLO (not necessary to publish - its working on the new node of the new subnet)
eth0 202.130.34.115 snm 255.255.255.248 gw 202.130.34.113 bc 202.130.34.119
eth1 202.130.34.116 snm 255.255.255.248 gw 202.130.34.113 bc 202.130.34.119

So, lets focus on eth0...

We wanted to do a phased transition by binding the old and new IPs to the eth0 but I couldn't get the default routes to work so I figured I'll put up with the DNS outage while propagation takes place overnight and just switch over to the new IPs physically.

I did this by being able to use iLO as a fallback to get ssh access when I stuffed up; and configured the two ethernet adapters as above...

Now I can ssh in to the box via iLO, eth0 on 202.130.34.115 or eth1 on 202.130.34.116

I reconfigured named and apache2 to point to the new IPs everywhere I could see that it needed doing and for a while after complete outage of all the sites, the sites all worked ok on the new IPs! I was able to send mail via roundcube on my https domain, use the database on the primary forums domain on the new IP; I went to bed.
This morning and today none of the sites are working and although name resolution is working locally on the server nothing resolves from outside world.
_________________
]8P