hacmp 5 cascading fallback

[karim93]

Hi ,

there is my configuration :
2 nodes Aix 5.2 ml05, hacmp 5.2
2 network adapter on each node
1 serial network

node 1 hostname anna en0:192.168.45.100 en1:192.168.0.1
node 2 hostname maria en0:192.168.45.80 en1:192.168.0.2

ip labels:
192.168.45.100 anna
192.168.0.1 annastb
192.168.45.1 anna_service
192.168.45.80 maria
192.168.0.2 mariastb
192.168.45.2 maria_service

one ether network IP Address Takeover via IP Aliases
2 classic cascading ressource group

Startup: Online on Home Node Only
Fallover to Next Priority Node in the List
Fallback: Fallback to Higher Priority Node in the List


rg1 : anna maria service ip label: anna_service
rg2: maria anna service ip label : maria_service

everything works fine, on startup rg1 comes online one anna
rg2 on maria
halt -q on anna : ok rg1 goes to maria

re-start anna & cluster services rg1 fallback to anna

halt -q on maria: ok rg2 goes to anna
re-start maria & clusterservices rg2 fallback to maria

disconnecting one network cable on any node : ok, service ip moves to other adapter

now ....what i am wondering..
when I disconnect the 2 network cable on a node, the ressource is moving on the other node (as espected), (network_down_local , rg_move..., )

waiting for cluster to stabilize

I re-plug the 2 cables on the node, there is a "network_up""join_interface" on log, but no "rg_move" as i was supposed it should be, the ressources stay on the takeover node.

what do you think about it? is it a wrong configuration?
or it is just normal?

did someone do this test and had the same result?



thanks for your help

 

[DSMARWAY]

hi ,

if you removed both network cables from both cards i would see a network down , but the resource group not failing over to the secondary node because your serial heartbeat is still active.

This also comes under multiple failure and you have to be very unlucky to lose both network cards

 

[karim93]

hi,

i m not agree ! the ressource group fails over the secondary node when removing both network cables from the first, even if the serial heartbeat is stil active !

it is the normal beavior since hacmp 4.5

 

[DSMARWAY]

hi ,
thats interesting , because i would still see the server up and running and not failed over becuase the heartbeat via your serial interface is active and it can communicate with the other node successfully. Are you sure your serial heartbeat is working ?

The other point about resource group not failing back , this is normal because , you don't want the resource group failed back automatically once your network comes back up because you would have users on the 2nd node using the system , depending on your application , just failing it back without users not logging off could corrupt your data , hence the failback would be a manual task.

 

[karim93]

hi,

yes my serial hearbeat is working, if you don t have a good one in hacmp 5x, you have a "node isolation" when you re-plug network cables (the 2nd node stops automatically)

the normal beavior should be that the ressource fallback to the first node after a netork_up, i think, i am wondering why it doesn t fallback, no matter what "we" want
you know what i mean ?