Home » Server Options » RAC & Failsafe » Clusterware and second virtual interface (10g2, Centos 5.4 (modded))
Clusterware and second virtual interface [message #458267] Fri, 28 May 2010 05:53 Go to next message
generalacc
Messages: 4
Registered: May 2010
Location: Greece
Junior Member
Hello,
We are building a clusterware implementation to protect a oracle SINGLE instance DB (using the well known pdf

guide) and a 3rd party application (i'll call it APP). We want to assign 2 virtual IPs, one for the DB and one for

the APP. The fact is that we want to split the traffic to 2 ethernet cards by assigning IP to boths of them. The problem is that I cannot assign VIPs to the second network card. Some facts:

Oracle 10g2

oifcfg -getif:
eth0 192.168.0.0 global public
eth1 192.168.0.0 global public
eth2 192.168.3.0 global cluster_interconnect

crs_stat -t -v (truncated)
rg1 application ONLINE esm1
rg1.lsn application ONLINE esm1
rg1.vip application ONLINE esm1 << Bound on eth0 WORKS
rg1.db application ONLINE esm1
arcVIP applcation OFFLINE esm2 << Bound on eth1 problem

command used to create APP VIP:
crs_profile -create arcVIP -t application -a $CRS_HOME/usrvip -o oi=eth1,ov=192.168.0.10,on=255.255.255.0

trying to start bring the following vip log on both nodes:
Interface eth1 checked failed (host=esm1)
Invalid parameters, or failed to bring up VIP

I also cannot run vipca (various java library error-no time to debug a program too)

Is it possible to give a hint of what is going on? Is it possible that the same subnets pose problem?

Regards,
Jim
Re: Clusterware and second virtual interface [message #458277 is a reply to message #458267] Fri, 28 May 2010 06:54 Go to previous messageGo to next message
John Watson
Messages: 8922
Registered: January 2010
Location: Global Village
Senior Member
The same subnet is not an issue, though using 192.x.x.x for VIPs may (according to the RFC that describes VIPs) not be allowed and definitely causes problems with some Oracle 10g tools.
I would fix Java first, how do you know that some tools other than vipca do not need it? Then use vipca, which does not care about non-routable 192 addresses. And what did you do with the crs_setperm utility? Did you run it at all?

edit - added:
And of course, what is in the log files?

[Updated on: Fri, 28 May 2010 06:56]

Report message to a moderator

Re: Clusterware and second virtual interface [message #458343 is a reply to message #458267] Fri, 28 May 2010 13:25 Go to previous messageGo to next message
mkounalis
Messages: 147
Registered: October 2009
Location: Dallas, TX
Senior Member
I am sure you knew this, but wanted to make sure - you are using these instructions?

http://download.oracle.com/docs/cd/B19306_01/rac.102/b14197/crschp.htm#sthref734

What version of Oracle? What OS?
Re: Clusterware and second virtual interface [message #458349 is a reply to message #458343] Fri, 28 May 2010 13:52 Go to previous messageGo to next message
mkounalis
Messages: 147
Registered: October 2009
Location: Dallas, TX
Senior Member
And one other thing - CentOS is not something that Oracle has ever tested their software on, nor is it supported. I have personally been involved in a few escalations where an Oracle customer attempted to use CentOS believing it was an exact copy of Redhat, only to find out that there are in fact differences.
Re: Clusterware and second virtual interface [message #458365 is a reply to message #458349] Fri, 28 May 2010 18:56 Go to previous messageGo to next message
generalacc
Messages: 4
Registered: May 2010
Location: Greece
Junior Member
Thank you both for your interest,
yes, I am well aware of the order and meaning of all the commands and options (i am not a DBA though). The single instance database is
successfully serviced throught the eth0 interface which was selected as public during the clusterware installation. The second
application cannot be started as the application VIP cannot be started at first hand. Apparently, CentOS 5.4 where successfully
transparent during the cluvfy checkups. Oracle is 10g2.
Mr Watson, my log files do not show anything clearer than my [vip].log: Interface eth1 checked failed (host=esm1)
If you can guide through some other logging mechanism or trace file, I can bring you some more info. I know such problems are very
general in nature and I do not expect a straight forward answer.
Thanx again for you interest.

[Updated on: Fri, 28 May 2010 18:57]

Report message to a moderator

Re: Clusterware and second virtual interface [message #458366 is a reply to message #458267] Fri, 28 May 2010 19:51 Go to previous messageGo to next message
mkounalis
Messages: 147
Registered: October 2009
Location: Dallas, TX
Senior Member
Just a word of caution - cluvfy successfully running does not mean that Oracle will work - it's expected to be run on supported OS's - not everything. Still, the fact that it did work may mean you are OK - but there is a lot more behind the scenes going on that cluvfy doesn't check.

Can you manually plumb the second IP address to the nic? Are there any messages in the /var/log/messages?

What point release of 10gr2 are you running? (10.2.0.3, 10.2.0.4, etc.)

[Updated on: Fri, 28 May 2010 19:54]

Report message to a moderator

Re: Clusterware and second virtual interface [message #458386 is a reply to message #458366] Sat, 29 May 2010 05:22 Go to previous messageGo to next message
John Watson
Messages: 8922
Registered: January 2010
Location: Global Village
Senior Member
Did you use the crs_setperm utility to set the permissions on the VIP appropriately?
Or have you fixed Java, so that you can use vipca?
And what error is in the crsd.log file, when you try to start the VIP?
Re: Clusterware and second virtual interface [message #458387 is a reply to message #458386] Sat, 29 May 2010 05:45 Go to previous messageGo to next message
generalacc
Messages: 4
Registered: May 2010
Location: Greece
Junior Member
Yes crs_setperm was used correctly as owner:root and user oracle to have read&execute
proviledges. I am not at the production site at the moment due to weekend but as I remember
crsd.log does not provide something helpful, I would have spotted it.. I reproduced some parts
of the environment in VMWARE at home and the application VIPs are created successfully on the
second interface. I will search for differences in the configuration on Monday and report back.
Hope you are still here.
Re: Clusterware and second virtual interface [message #458621 is a reply to message #458267] Mon, 31 May 2010 13:21 Go to previous messageGo to next message
generalacc
Messages: 4
Registered: May 2010
Location: Greece
Junior Member
I managed to spot the source of the problem. It lies in the racgvip shell script where a ping is done to the default gateway
in order to determine if the interface is up. The problem was that the firewall was blocking the ICMPs thus an error code
is returned and the interface is marked useless. For the time being I comment the "ping" line (as I know the interface is up)
resolving the issue. Have to check with the net admin tomorrow.

Thanks both of you for your time and interest.
Re: Clusterware and second virtual interface [message #458624 is a reply to message #458621] Mon, 31 May 2010 14:25 Go to previous message
John Watson
Messages: 8922
Registered: January 2010
Location: Global Village
Senior Member
Excellent catch. I've just looked at the script. A nice example of a piece of code that was written with the best of intentions, but can have an ludicrous effect. Thanks for sharing this.
Previous Topic: Fatal cluser error encountered....
Next Topic: Planning For RAC on 11g2
Goto Forum:
  


Current Time: Thu Mar 28 06:44:13 CDT 2024