server

[David_492]

Dear all

I have one HP UNIX server rp8400 i already partioned into two hardware (npar) hpnodes hpnode1 and hpnode2 and i made the clustering between them and i defined the package my question how could i test that the clustering is working well in case of one of the nodes is failing over or the ethernet card of that node is down?

2.If i applied apache server over these two nodes how i can test the clustering?

3. if i defined logical ip address over the ethernet cards which are related to those two hpnodes also the apache server is applied there how i can test the ping command incase one of ethrenet is down over one of these nodes or in case of any failover?

Thank you all

[Ashwani Kashyap]

1. Pull out the network cable from the primary lan cards and see if the lan fails to standby by pinging the IP from another host . First use cmhaltpkg to halt a package and use cmrunpkg to ru it on the alternate node . Then finally you can halt one npar and see if all the packages failover to the other npar .

2. Same way as described above if apache is in a package. or you can run individual instances of apache on both the nodes therby reducng the number of packages .

3. Ping the logcal IP address form anywhere n your network , before failing . Then failover and ping it again . If pingable , then it has failed over correctly . YOu can laso try telnet .

[Joe Short]

1. To get a more complete look at your test, log onto one of the systems. First run "cmviewcl -v" on the system you are logged into. On the other system, unplug the primary NIC. Then run "cmviewcl -v" again. This SHOULD show that the IP has moved onto the alternate NIC.

2. If Apache has been installed on the package and not on one of the nodes, it should fail over from node to node. You should also incoroprate the proper Apache start and stop commands into the package.cntl
If the package fails from server to server, the processes should be started on the new server.

3. Use the "cmviewcl -v" command to verify that the package has failed over. if the PACKAGE IP answers your ping the test is a success.

P.S. I would be very hesitant in implementing a Service Guard cluster on nPars in a single server. Althnough it can be done, it does seriously limit the High avaialability capabilities of MC/ServiceGuard. If for any reason the server itself should go dark, then the entire cluster is dark, and you have no failover. You want to remove as many Single Points of Failure(SPOF) as possible.

Joseph M. Short
Senior Technical Consultant
AdvizeX Technologies