RPC question

[made13]

Scenario: 2 AIX P630 servers.

NodeA has multiple ip addresses on single interface.
Primary ip address on en1 is the boot-address. Aliased to it is a service address, serviceA.

NodeB makes RPC calls to NodeA, on the service ip address serviceA. However, nodeA replies with the boot address of the machine and not the service address.

How do we get nodeA to reply on the service address rather than the boot address?

 

[MoshiachNow]

HI,

you have to define statis routing on the machine with the multiple interfaces - NodeA.
Also have to define the proper NodeA interface address in the /etc/hosts of NodeB.

See the following link for the static routing reference:

http://publib.boulder.ibm.com/doc_link/ko_KR/a_doc_lib/aixbman/commadmn/tcp_route.htm#A9C0aa54

Long live king Moshiach !

http://www.7for70.com/

 

[lillracksingen]

Also:
1. Make sure that your RPC server on nodeA has bound to the right IP-address. Use [tt]rpcinfo -p ipaddress[/tt] to see socket-number and then [tt]netstat -anf inet|grep socket[/tt].
2. You may need to delay the start of the RPC server on nodeA until after configuring the service-address. If this is an HACMP cluster then start your processes by using run-level a in [tt]/etc/inittab[/tt].

 

[made13]

Thanks for your replies. NodeA is part of a HACMP cluster running CASCADING resource groups and IP aliasing.

Question is how do we delay the start of the RPC server on nodeA until HACMP has started up. Looked at /etc/inittab, and I can't see any of the entries with run level 'a'. Note that this cluster is running HACMP 5.1 on aix 5.2.

Compared this to a cluster running HACMP 4.4.1 on aix 4.3, and I can see rctcpip, rcnfs, qdaemon, writesrv, clinit, pst_clinit all runninf at run level 'a'. Is this something you may have come across due to differences in the versions?