strange behavirs of hacmp (5.2)

[ogniemi]

We have a simple cluster with two resource groups - each one taking care about 2 service IP addresses (so in total 4 service addresses in the cluster).

There are two cluster nodes with 4 network inetrfaces per node (no persistent addresses) IPAT is made by addres takeover (I am not using IPAT via aliasing).

So I have 2 boot ifs and 2 stby ifs per node.

Each resource group "has" its own home node, startup policy is set to start on first available node.

So, when I start cluster on first node, both rgs are bring up smoothly on that node. I have then 4 service addresses configured on that node (no stby left). At this state, netstat -r shows:

rg2_serv loopback UGHS 15 2879555 lo0 - - -
rg2_serv2 loopback UGHS 15 2343465 lo0 - - -


When I do start HACMP on second node, (the high prorioty node for rg2), then rg2 is released (stopped) on first node and started on second node, but...the routing entries mentioned above are not removed from first node by HACMP!!!!!!

Services rg2_serv and rg2_serv2 are up on node2 but I cannot ping from that node rg1_serv or rg1_serv2 being up on node1 !!!!!!

To solve this problem I run on node1:
route delete -host rg2_serv loopback
route delete -host rg2_serv2 loopback

but shouldn't it be done by HACMP itself / during takeover?


(HACMP 5.2 with latest fixes and some exifes, AIX 5.2)

 

[khalidaaa]

I beleive HACMP should run on AIX 5.3!

Regards,
Khalid

 

[ogniemi]

why?

have you any reference, url saying it?

 

[ogniemi]

I am sure the OS is OK.

also hacmp 5.3 can be instaled on AIX 5.2

Question: On which AIX 5L levels is HACMP V5.3 supported?

Answer: HACMP V5.3 requires one of:

AIX 5L V5.2
AIX 5L V5.3

HACMP V5.3 will support future releases of AIX 5L as they become available. To find the latest fixes available for HACMP and AIX 5L, refer to the TechSupport web site.

 

[khalidaaa]

well, i'm not really sure with that

coz we had a project of creating three clusters on P5 570 machine and i think one of the reasons to go for 5.3 was that HACMP is not avaiable on 5.2!!!

but let me confirm it and come back to you

Regards,
Khalid

 

[khalidaaa]

Oh i'm really sorry my friend

I confused two things DLPAR and HACMP

I'm soo sorry

 

[norestore]

Hi,

well, I don't have really a solution for you...

The handling of the routes thru the takeover is done by the event-Script "cl_swap_IP_address". This is part of the cluster.es.server.events fileset of HACMP 5.2.

Check it out.

When you are already at the actual levels of AIX and latest HA-fixes, perhaps you have to make a supportcall towards IBM.

Regards, Martin

 

[mrn]

The route should be created by ifconfig when a swap_adapter event occurs.

Though the HACMP script has been working across several
different versions and levels of AIX and HACMP through
use of deleting the route and ifconfig down of the interfaces,
at some level of change within AIX this is no longer
consistent. As recommended by AIX TCP I have changed the
cl_swap_IP_address script on my nodes to detach both interfaces before ifconfig either of them back up.



Mike

"A foolproof method for sculpting an elephant: first, get a huge block of marble, then you chip away everything that doesn't look like an elephant."

http://www.airport-parking-site.co.uk/