[aix_tiger]

Guys!

I need your advices for setting up network for HACMP implementation ( in dual VIOS environment, dual P550 servers)....

I do have some questions and appreciate , if somebody can response on these...

First , as i do have two physical ethernet adapters on each VIOS, i will etherchannel them ( for load balancing and HA)... and then configure SEA failover on both VIOS servers... Questions are...

1. Can i assign ip address and host names for VIOS servers on SEA adapters, as follows?

192.168.10.1 ---> 192.168.10.4 ( VIOS1, VIOS2, VIOS3 and VIOS4)

Please note that HMC IP address on ethernet interface ( communicating with client or public LAN ) will be 192.168.10.5

2. All client Lpars ( six on each p550 ) will have one virtual ethernet interface each.......and then configure HACMP with single nework interface ( Is it mandatory or i can go ahead with two adapters per cluster node , as we usually do in normal HACMP configurations?)


3. Boot Ip addresses ( assigned to client virtual ethernet adapters ) will be 10.100.101.20 and 10.100.102.20 , while service IP address would be 192.168.10.6 ( cascading single Resource group with failback configuration )... is this seems OK ?


4. can i distribute virtual ethernet traffic for lpars created across two P550s , using different VLANs? Means that if i do have one Lpar ( say lpar1)from p550-1 box and other lpar ( say lpar2)from p550-2 , being configured as same cluster nodes , can i assign virtual ethernet cards on lpar1 and lpar2 with different vlan IDs ( mapped with one SEA on VIOS1 on p550-1 and other with SEA on VIOS1 on p550-2)... Or it is mandatory to have same VLAN IDs assigned to virtual ethernet interfaces partcipating in same cluster?

Thanks in advance for help....

[Jorke]

1) yes you can, please be aware that this is not the recommended IBM configuration, they recommenend a seperate card for the management access to the VIO server.

2)HACMP doesn't require a second ethernet adapter and if you use ip aliasing it will work. (if you use the adavanced config) But no Ethernet adapter failover and you will get warnings in the verification. And it is a small effort to assign a second card to an LPAR. My recommendation would be to assign a second one to remove these annoying warnings.

3)if you can post the complete /etc/hosts file it would be more clear. but this configuration should work.

4)the two ethernet cards needs to communicate with eachother if configured in HACMP. If they are in different VLAN.. they will think they are down. and that ain't good. So no two VLAN's for HACMP.

Regards,

Jorke

[aix_tiger]

Hi Jorke,


First of all many thanks for your kindl reply...

Your answers has given rise many more questions in my mind...Let's share with all , one by one:


1. You said that IBM recommend a separate management LAN... but for what? Are you refering towards management LAN of HMC itself, if yes then off course HMC management LAN using DHCP would be on seprate network.. while only one of its interface will be communicating to VIOS servers .. please explain further.

2. I agree with you that this kind of setup will work with single ethernet interface HACMP networks easily , however with two interfaces on each HACMP node will add complexity.. In that case , i would be ending up with two virtual ethernet interfaces on each cluster node , but what will happen to VLAN IDs??? Should I , in that case, create all of these virtual interfaces on client LPars in same internal VLAN with VLAN ID =1?

Thanks again!!