A question about TCP/IP

[olive_wide]

I have two Alha server ES45 nodeA and nodeB,same config(hardware and software).
I want to make a programme to achieve the following functions.
NodeA and NodeB are running.When nodeA is shutdown,run this programme at nodeB,then the nodeB's tcpip config restart,it BECOMEs nodeA's tcpip config,includes HOSTNAME ,IP ADDRESS,Service,proxy,route,etc.
My idea:I can copy nodeA's tcpip database to the nodeB,then,before restart tcpip of nodeB,I can CHANGE the system PATH loading these databases.I don't know if I can change this paths,and how to do.Or you have anothe idea to achieve this function?
Thank you very much !
olive

[Uwe Zessin · Accepted solution]

There is a lot of risk involved when copying a few configuration files around and this is not enough for TCP/IP. You can easily leave a system in a terribly inconsistent state.

Can't you simply move the system disk drive from NodeA to NodeB?

[olive_wide]

UWE,
Thank you for your reply.
You say "it is not enough for tcp/ip".What else can I do? I tried to change the logical name like sys$startup, sys$system,and change this logical to another path,but failed.
I know that it is very dangerous for the system,but can you give me another idea?
NOT MOVE the system disk,only restart TCP/IP.

[Antoniov.]

Olive,
I understand you want create a backup machine. For this purpose you can use some advanced solutions like clustering
http://h71000.www7.hp.com/wizard/wiz_2510.html
or network technologies
http://h71000.www7.hp.com/openvms/journal/v2/openvms_journal/css/openvms_journal.html

Copying files may be very dangerous.

Antonio Vigliotti

[Antoniov.]

Olive,
here there are tcpip files from hp documentation
http://h71000.www7.hp.com/doc/82FINAL/6524/6524pro_002.html#sec_creating_new

Good Luck

Antonio Vigliotti

[olive_wide]

Antoniov,
Thank you.
Yes,I want to make a backup function.
My idea is I can copy the nodeA's tcpip information(including IP ADDRESS, HOSTNAME, SERVICE,etc),and put these informations in a SPECIAL path in nodeB.When I want to use it,I will CHANGE SOMETHING,then restart the tcpip,then OK,the tcpip config will become nodeA.
I want to know how to achieve this function,although it is very very dangerous!
olive

[Kris Clippeleyr]

Hi,
As others suggested, you might consider building a cluster (two nodes with a directly accessible quorum disk should suffice).
Then, instead of copying IP configs from one system to another, you could define an IP alias that is bound to one machine at the time. If let's say node A has the alias, and thus running your app, and then, for whatever reason, disappears from the cluster, you could then give the IP alias to node B, and run the app on B. Please check the IP documentation on how to configure an IP alias.
Regards,
Kris (aka Qkcl)

[Jan van den Ende]

Olive,

those who know me will also know that I am all for the cluster solution!

But, IF there are reasons at your site to NOT do that, then maybe you can look at this from another side:

This is, assuming you ARE using DNS, and your DNS server is NOT nodeA nor nodeB.

Of course you have DNS names for each node.
Now, do NOT set up your clients to use those names DIRECTLY, but, for each (group of) application(s), set up a name reflecting the function as a CNAME to the node where the application is available.

Now, in case of failover, -- in the DNS -- you change the translation of the CNAME(s), and that is ALL you will have to do to fail over the connectivity concerned. Note well: There may well be authorisation and data issues as well.
In that case you REALLY should go the cluster way, because that does not just solve such issues, they just ARE NOT THERE, because of the single system view of VMS clustering.

hth,

Proost.

Have one on me.

jpe

[olive_wide]

Jan van den Ende
Thanks for your reply.
I think I have to change my idea,because all of you suggest me to use cluster solution.
And I will think your another advice.
The safety of the system is the most important thing.
thanks all.

olive

[Anton van Ruitenbeek]

Olive,

What I don't know yet (or I missed it somewhere) is the versions you are using. If you create a cluster of these nodes you can add also FAILSAVE_IP to it (5.4 eco 4). If you do so and one of the systems is dead, the other system automaticly takes over the IP etc from the down system. If you are using it as a web-server and the system is getting up, the system gets his own IP-adres back. Otherwise if there are connections open it will not automaticly takes his setting back. Then you have to do this manualy because of the open sessions has to be discarded. But that isn't such a big deal. Just go into ifconfig and do '-fail' the ip adres to go back.
IMHO never try to stop IP and start with settings from another node. A lot of 'underlined' stuff has to be changed as well (nodenames in HOST table, in de TCPIP CONFIG environment etc.) I do it sometimes and even whith me it doesn't come up at once, I always has to add some stuff somewhere where I keeps on forgetting it. Its not safe to do so !!

AvR