Changing IP address

[Phil.Howell]

I need to swap a 4-node vms alpha cluster + a standalone development machine to a new network
All are running vms 7.2-1 and tcpip serv 5.1
Can all the tcpip config changes be done using sys$manager:tcpip$config.com? I have used the tcpip command line interface to do minor changes in the past, and I usually get it right at about the 3rd attempt, so I thought that if I built (and tested) a sequence of tcpip commands on the standalone machine to do these changes, then I could then run the same procedure on each node with the ip numbers changed.
Any recommendation on which way to do it?
(decnet IV is also used but I am assuming that this will just work as usual)
Phil

[Kris Clippeleyr]

Phil,

Am I correct in assuming the nodes will get new IP-addresses? If so, simply execute @SYS$MANAGER:TCPIP$CONFIG and select "1 - Core environment", then "2 - Interfaces".
You can then (depending on your hardware) reconfigure (i.e. give a new IP-address etc.) the Ethernet device.
You need to restart TCP/IP to make the changes effective.

Kris (aka Qkcl)

[Ian Miller.]

yes you could have a sequence of commands.
$ TCPIP SET name/ADDRRESS=a.b.c.d
$ TCPIP SET CONFIG INTERFACE xxx/HOST=name/NETWORK_MASK=w.x.y.z

You may need to put name in "" if its currently lower case?

Is the domain remaining the same?

[Robert Gezelter]

Phil Howell,

Two notes:

If you are careful, you can do this when logged in to the cluster, using the DECnet transport (SET HOST or SYSMAN), even if your connection from your desktop is via IP.

Also, be careful about the DNS change, it can take time to propogate. If you are available through an internal DNS, consider putting an alternate network adapter on the new IP address on one adapter and leaving the other on the old network for a changeover period, to deal with the changeover.

Otherwise, if you are using some form of cluster load balancing, consider switching the cluster in two steps, port side (2 machines) on one night, starboard (2 machines) on the other night. In that case, the combination of the IP address switch and the DNS change will interact to produce a temporary period where two of the systems are temporarily unreachable (until the DNS caches refresh), but there is no need for an interruption in service.

I hope that the above is helpful.



- Bob Gezelter, http://www.rlgsc.com

[Volker Halle]

Phil,

I would do this by first removing the IP interface and local hostname from the local TCPIP configuration database and then run @TCPIP$CONFIG to define the new interface IP address (and host name).

$ TCPIP SET CONFIG NOINT xx0
$ TCPIP SET NOHOST ip-name

Otherwise the TCPIP configuration procedure will complain, that there is already a host with the same name, if you enter a new IP address and try to use the same local host IP name.

Volker.

[Jan van den Ende · Accepted solution]

Phil,

we had to do the same thing.

We more or less worked along the lines Bob advises, in using an alternate adapter. We did not do two nodes at a time though. Main reason: still allow quorum to be maintained should one node fail.

Maybe we had some advantage: our requirements for redundance demanded that we also have double networking, so each machine was to get an extra NIC anyway.

We disabled all new service connections to one node. When users were all gone, we installed the second NIC (and used the occasion to install patches). Since you can not have two cards with the same name, we included the device name into the IP adres per card. That means we also had tho change the DNS: Now the node-name is a round-robin CNAME over the node's node-NIC names.

After all name caches had had time to flush their contents, the services were re-enabled for the upgraded node, and the next node was disabled. .

The users never noticed.

Success.
Do report your results, please.

Proost.

Have one on me.

jpe

[Phil.Howell]

Thanks for the replies, additional info: the domain is remaining the same, but we are moving from a dec switch to a cisco, and changing the current ip addresses to 10.n.n.n in a vlan configuration, so are also changing gateways and dns servers. The cluster also has a "cluster alias" ip address.
I will look at the replies in more detail and probably do the changeover this weekend.
Phil
OT
On the subject of "proost have one on me", I can buy 330cl bottles of "Mariner premium lager, brewed and bottled in Amsterdam" for $AU 1.25 (which equates to 51p or about .75 euro) in the local supermarket. How can they ship this halfway round the world, and then the 500km from Sydney for this price? Is this another EEC surplus that is being foisted on the rest of the world, or just one of the benefits of globalisation? How much do you pay in Europe?

[Wim Van den Wyngaert]

Phil,

I don't know about Holland but in Belgium the major part of the proce of a beer is taxes. Just as all petro products. So, comparing prices is difficult.

In Belgium the costs are (currently in Makro) :
6 Kriek beers : 0.62 EUR
24 pils : 7.81 (so a lot more than in AU)
1 can of pils : 0.61 EUR(almost 6 times your price)
25 Leffe Brown : 12.63 EUR

Wim

[Anton van Ruitenbeek]

AFAIK

@Wim,

Yes, in the Netherlands the major part is also taxes.

@Phil,

It can be localy brewed. Normaly all the liguid is shipped in containers and bottled localy. And by ship the cost are only per volume and not by weight.

AvR

[Jan van den Ende]

Phil,

...and if you really want it fance, go to Sandinavia. Especially Sweden is 'nice':

_ANYTHING_ containing more then 3.5 % v/v ethanol cam _ONLY_ be sold in state "shoppes".
To me, they best comparison I could find were the move caricatures of Sowiet food shops in times of scarseness!.

Already some years ago, I did not pay the EUR 6.-- for .3 L of 5% lager...
because we had our mobilhome weighed down with can of a special offer from a Dutch supermarket :-)
Image the prices for the stronger, and the real strong, stuff!

This really makes it understandable why "every other Swede outside the cities" has has own still.

Proost.

Have one on me. (but not Swedish-bought!!)

jpe