MCSG and Oracle Problem

[Jayan_2]

Dear MCSG Admins,

I have a problem with a two node cluster.
Nodes are K370 with 2X 100BT lan out of which one is standby.Heart beat on inbuilt 10MBPS card.I have two packages configured,one in each node.Os version is HP UX 11.0,ORacle is one package and PDM ( A X windows CAD Application ) is another pakage.Both these packages got floating ips.Issue is only with the oracle pakage,the package is running in the original node always,but every five minutes there is a connectivity problem.When you ping from the clients to the oracle pckagee IP for 5 minutes it pings and another % minutes it says request times out.How ever we are able to ping the server where the oracle package is running.

Any insight on this issue.

If you could throw some light on this it will very helpfull.

TIA

Jayan

[Geoff Wild]

Wow - that's pretty wide open....

Things to check:

syslog
cluster logs
check for cron entries (all users) that run say every 5 minutes...
check network errors - netstat -in
make sure network is configured correctly - IE - is it 100 Mbps but downgraded to half duplex when the hub/switch port is set to 100 Fdx?
Are you running Oracle Nameserver? if yes, get oracle dba's to check logs...
What about Oracle Archiving? Do you archive to tape when a filesystem reaches a certain size? Maybe the bandwidth is being taken up by your backup server - if yes - then get another interface and create a "backup" network...



Rgds...Geoff

[Duncan Edmonstone]

Is there a firewall between the clients and the cluster? Firewalls can sometimes consider IP addresses with changing MAC addresses, or different IP addresses with the same MAC address as 'suspicious' and block them. Check your firewall logs.

HTH

[Helen French]

I would check the following:
1) Any errors in the package configuration script. Check for any timeout values.
2) Patches. Apply the latest MC/SG patches to system.
3) Any message from EMS about a possible H/W problem. Check root mail.
4) Check if the package trying to switch to the standby card during the time you are not able to communicate.
5) Check whether the package works fine on the other server. You can test it by a failover process.

[Ionut Grigorescu_2]

Hi,

check for errors in /etc/cmcluster//
thank you,
ionut