hacmp & adding static routes

[wab62]

Hello,

Having a problem with aix 4.3.3 and hacmp 4.3.1. When cluster services are up, if i add a static route thru smit,
i lose my service adapter, it appears to be down. The boot adapter then becomes active for the affected network. My nieve question is, should this work this way? Should this be done with cluster services down? Or is there some other method for adding static routes in an HACMP cluster? Thanks for the help.

 

[usige]

wab62,
In our config we store all static routes in an rc file called rc.routes. The routes are added as part of the start script for the resource group.

I don't know if the behavior you describe is normal. As a general rule I find that mucking with the network via non-cluster smit panels does not always produce the expected results. Having the routes added after the service adapter is up is as easy as a list of route add .... commands. This is the reason we have them in an rc script, as routes added this way don't interact with ODM and are therefore not persistent.

Hope that helps.

 

[ademeter]

Hi,
you loose your service address when HACMP detects a network/adapter failure and it brings the resource group down, eventually up elsewhere. Adding routes through "smit route" should work.
There may be a problem with the routes if you are adding host specific routes those conflict with the HACMP failure detection logic. HACMP changes the route tables on HACMP nodes to diferentiate network failures / adapter failures / node failures.
However, I have met strange behaviour on older HACMP versions (I think it was 4.3.1) when HACMP brought down the service adapter after detecting network_down event, but the resource group stayed online.
I would recommend to double check the IP (ifconfig -a), routing (netstat -rn) and HACMP (cllsif) konfiguration. It is worth to check the "no" attributes and focus on ipignoreredirects=0.

 

[wab62]

using the route command worked ok, so i simply added them to rc.net.

thanks for the help.

 

[leocst]

Dear all
I got the same problem on the production server.

how can i fix it within a minimal down time? or i need to stay in computer hall from christmas eve to boxing day ?

need to uninstall and reinstall HACMP or there is any much simple way?

best regards
Leo