How to specify which adapter nrglbd should use
ITEM: RTA000086348
Using AIX 3.2.5 and NCS in a machine with more than one
network adapter, I want to tell nrglbd wich adapter to use.
I seems to me that nrglbd will choose the adapter first
configured, and the only way to change this is to delete
the wrong adapter and create it again. This will work if
you are in a static environment, but if you are using
HACMP you can never trust, that nrglbd will start up on the
right adapter.
So I need a way to tell nrglbd which network adapter to choose.
We are running an Imagesystem called ODIN which is based on NCS.
The system are running on a RISC/6000 with AIX 3.2.5 and 2
Token-ring adapters called prodsrv. On another RISC/6000 also
with 2 token-ring adapters we are running an ODIN testsystem
called testsrv. The 2 machines have both HACMP installed, so
testsrv can takeover the production if prodsrv fails.
The 2 token-ring adapters have their own hostname and ip addresses.
The nrglbd will use one of them, but you can never tell wich one
because the lag of a flag or file to tell nrglbd wich network
interface to choose. As mentioned earlier, it seems that when we
start nrglbd it will take the 1. adapter configured i the ODM database.
That what we found out after we restored another Imagesystem from
tape and unfortunately configured the adapters in the wrong order.
But HACMP do reconfiguration every time a shift from the prodsrv to
testsrv occours. The llbd's are configured to a certain hostname or
ip-address in the glb_sites file, but we mis a similar facility to
tell nrglbd wich hostname or ip-address to use.
We are considering a circumvention, to define the 2 possible
ip-adresses in all the glb_sites files, hoping that the llbd in this
way will find the address of the nrglbd whereever he are.
The right solution would be a flag -i: startsrc -s nrglbd -i tr0
to tell nrglbd always to use tr0.
ANSWER
Although it is not said specifically, I believe from your information
that you are running NCS 1.0 in your configuration. A first comment is
that it would probably be better, if you are running in an HACMP cluster,
to be running NCS 1.5.1 instead. This gives you the ability to run the
glbd, which is replicatible and synchronized between servers, instead of
the nrglbd, which does not have this ability.
The way to specify which adapter to be used by the global location broker
is to detach all the interfaces that you do not want used before doing
your NCS configuration, then reactivating them again. In the case of
HACMP, you would want to detach your standby adapter, with:
# /usr/sbin/chdev -l INTERFACENAME -a state=detach
or
# /etc/ifconfig INTERFACENAME detach
Then do your NCS configuration, then reactivate your standby adapter:
# /usr/sbin/chdev -l INTERFACENAME -a state=up
There is a manual I would highly recommend for guidance with this and
related concerns, called iFOR/LS Tips and Techniques (SC23-2666). It is
written for iFOR/LS setup, which incorporates NCS as well, and gives a
description, and configuration shell scripts, for setting this up to work
in an HACMP environment. The suggestion above is from this manual.
S e a r c h - k e y w o r d s:
NCS HACMP INTERFACE NRGLBD GLBD STANDBY
WWQA: ITEM: RTA000086348 ITEM: RTA000086348
Dated: 11/1995 Category: ITSAIHA6000
This HTML file was generated 99/06/24~12:43:28
Comments or suggestions?
Contact us