FailsafeIP Reconfigure

[Jack Trachtman]

VMS V8.3
FailsafeIP

I have a node with two ethernet ports, one configured as primary the other as standby.

I would like to change the definitions of which is primary and which is standby (ie there will be no IP address changes, just FailsafeIP changes.)

Is there a way to do this while keeping the system on-line, ie w/o having to first deassign all IP-to-port address assignments and then redefine them?

Thanks

[Jon Pinkley · Accepted solution]

Jack,

Please do the following:

$ @sys$manager:tcpip$define_commands.com

Then copy/paste the output from the following into your next response:

$ ifconfig -av

Also provide exactly what you want the configuration to be after the change.

What were your goals in using FailSAFE IP. Did you also consider LAN Failover (the LLA virtual device)

It isn't clear to me what you mean by "primary". Failsafe has active and standby, and "home" interfaces. Do you mean home or active?

Suggested reading, in addition to the documentation in the TCP manuals:

Matt Muggeridge - Configuring TCP/IP for High Availability

http://h71000.www7.hp.com/openvms/journal/v2/articles/tcpip.html

HoffmanLabs IP Tips: OpenVMS Clusters and Load Balancing

http://labs.hoffmanlabs.com/node/859

There are many other references you can find with a Google search.

Jon

[Jack Trachtman]

Here's the output requested (sorry that it wraps)

*TSTLAB::USER>ifconfig -av
IE0: flags=c43
*inet 10.4.32.140 netmask fffffe00 broadcast 10.4.33.255 ipmtu 1500 (Standby on TSTLAB)
HWaddr aa:00:04:00:02:04

LO0: flags=100c89
inet 127.0.0.1 netmask ff000000 ipmtu 4096

TN0: flags=80

TN1: flags=80

WE0: flags=c03
Standby addresses:
WEA0 inet 10.4.32.140 netmask fffffe00 broadcast 10.4.33.255 (Active on TSTLAB IE0)
HWaddr 00:10:18:10:c6:a0


Presently IE0 is Active & WE0 is Standby. Since the IE device is 100Mb & the WE device is 1Gb, I'd like to switch them.

[Jon Pinkley]

Jack,

Some additional useful information would be output from:

$ set proc/priv=all
$ show net
$ tcpip show version
$ mcr lancp show configuration
$ anal/system
SDA> show lan ! this will show what protocols each interface is using. Add /FULL for high volume output (mostly useful for debugging).
SDA> exit
$

Since the IE0: device has an active HW address that appears to have been set by DECnet IV, do you want the DECnet traffic to continue to use the IE0: interface? What do you want to happen if IE0 fails?

How was the WE0 interface configured?

If you haven't already done so, please read "HP TCP/IP Services for OpenVMS Management", "Chapter 5 Configuring and Managing failSAFE IP", and Matt Muggeridge - Configuring TCP/IP for High Availability

http://h71000.www7.hp.com/openvms/journal/v2/articles/tcpip.html

I don't have a similar setup, and I would have move a wire on a test machine (and I am not phyically where the test machine is right now) to be able to test it.

Here is what I can do on the test machine, which currently has a single interace that was configured with TCPIP$CONFIG.COM

I can remove the "home" attribute from the active connection without affecting connectivity using ifconfig -home

See attachment for the demonstration.

The manuals don't have as many examples as I would like, and I don't see anything that explicitly answers your question.

So I would guess that the following would work, although if it is a production system, I wouldn't try during a time that would affect users.

I would also make sure I had a secondary way to get into the system in case things don't go as planned. Ideally you should try this on a test system, or during scheduled maintenance period.

Ok, now that all the disclaimers are stated, I think this has a good probability of working:

$ ifconfig ie0 -home 10.4.32.140 255.255.254.0 ! this should remove the "home" attribute from the ie0 interface
$ ifconfig we0 home 10.4.32.140 255.255.254.0 ! this should make the WE0 interface the home interface, and should cause the active interace to switch

If WE0: does not become the active interface, then

$ DEFINE/SYSTEM TCPIP$FAILSAFE_FAILED_IE0 1

should cause the IE0 interface to switch to the WE0 interface the next time the interace counters are read by the failsafe. After the switchover you should deassign the FAILSAFE_FAILED_IE0 logical name, and the connection should stay on the home interface.

The following thread also has a similar question.

http://forums.itrc.hp.com/service/forums/questionanswer.do?threadId=1324128

You should also consider LAN Failover, which may be a better choice for your situation, where there is only a single IP address involved, and there are multiple protocols using the interface. For instance, LAN Failover will protect your DECnet IV traffic too. However, it only detects loss of link status; it isn't detecting loss of traffic. In other words, it only protects the connection to the switch or hub that the NIC is connected to. Failsail IP protects based on lack of packets arriving at the controller, which has its advantages and disadvantages.

You can actually use both LAN Failover and FailSafe IP at the same time, by putting the NICs in LAN failover sets, and having TCP use the LLA0 devices (which tcpip calls LE devices). But then you will need more NICs and possibly more "LANS", since I am not aware of a way to have two ethernet interfaces in use by DECnet IV on the same ethernet network, since DECnet IV changes the MAC address when it starts, and all ethernet interfaces used by DECnet IV will be set to the identical MAC address.

http://forums.itrc.hp.com/service/forums/questionanswer.do?threadId=907611

Another thread with LAN failover info

http://forums.itrc.hp.com/service/forums/questionanswer.do?threadId=1035709

Jon

[Jack Trachtman]

Joe - thanks for the info. It's useful in understanding where FSIP stores its config info.

Didn't know about the "home" option to ifconfig (it doesn't show up in "ifconfig -h")

Asked HP for suggestion & tried their idea. It didn't work and hosed TCPIP, but fortunately the system wasn't being used at that time.

Since the system was down anyway, I just deleted and recreated the FSIP config and now have what I need.

Thanks.

[Jon Pinkley]

Jack,

In case someone else has the problem you experienced, can you tell us what not to do, that "didn't work and hosed TCPIP"?

And I mean something other than "don't believe what HP support tells you and follow it blindly on a production system" :-)

When you reconfigured FSIP, what did you do differently than the first time?

Have you tested it?

I would consider the following a minimal test:

1. Reboot and verify that the configuration is the way you want (ifconfig -av)

2. Telnet into the protected ip address. Generate output traffic to the terminal (DCL loop printing something works)

3. Disconnect the active interface.

4. Verify that your telnet session is still working. Their will be a pause, but your connection shouldn't drop.

4a. I would expect your DECnet applications to fail at this point. Also any other protocol that doesn't have HA (high availability) built in.

5. Reconnect your preferred interface. The new interface should become the active one within a minute or two.

Jon

[Jack Trachtman]

Jon,

Some answers:

- the host originally had two 100mb ports which were configured to FSIP
- a Gb port was added and because we only have two physical enet switch connections, one cable was moved from a 100Mb port to the 1Gb port
- the IP port config was changed & the Standby def was moved from the 100Mb port to the 1Gb port, making that one the Standby node
- that's when I raised my question

The suggested HP solution (which left IP not working at all for some reason), is:

1. Define the logical $ define/system tcpip$failsafe_failed_ie0 1
2. Shut down failsafe $ @sys$manager:Tpcip$failsafe_shutdown
3. Deassign the logical $ deassign/system tcpip$failsafe_failed_ie0
4. Define the address back onto ie0 $ tcpip ifconfig ie0 alias

/
5. Restart failsafe ip $ @sys$manager:Tcpip$failsafe_startup

In config
1. Remove interfaces $ tcpip set config nointerface ie0 $ tcpip set config nointerface wea0
2. Go into tcpip$config and set up we0 and then ie0