HANA HA test cases

[Former Member]

Hi Folks,

We are in the process of testing HA on hana appliance.

We have 4 plus 1 node.

We dont have data in the appliance but we want to test HA part without any data first so that we have some benchmarks.

Now we have the below test cases like :

HDB stop on the master and slave

killing nameserver on master and slave

killing indexserver on master and slave

Any other options which we can think of ?

what impact does it have if we kill sapstart and sapstartsrv.

Message was edited by: Julie Blaufuss

[Former Member · Accepted solution]

Maybe you can add a scenario, pull out the network cable directly.

[Former Member · Accepted solution]

Wenjun,

Anything apart from physical, any other test scnerios you can think of?

[Former Member · Accepted solution]

Any other inputs?

[Former Member · Accepted solution]

"rm -rf *" on shared filesystem or on .ini files ? May be to hard...

fs saturation on one node (but it's means you have data)

activate iptables for a node to simulate a network break

change ulimit -m for <sid>adm to simulate memory leak on a node

[Former Member · Accepted solution]

Hi Romain,

Will the ulimit change affect memory leak since i think we need to restart right before it starts.

What all did you test?

Simulate a network break is a good idea, also how long did it take for your hana node to come up? since it has to load all the data into the new node right.

[Former Member · Accepted solution]

no, you are right, ulimit will not take place immediatly but you can stop the sid on one node of the system. set the ulimit in et /etc/security/limits.conf for your <sid>adm user and restart te sid on the node.

It is just ideas, no real tests for instance.

- use quota in gpfs for /hana/log and /hana/data and let files systems fill to the quota could be an other good test

[Former Member · Accepted solution]

Ferraton,

Thanks a ton for the inputs.

Did you try out any of the steps which you had mentioned?

we are not with IBM.

So have to use other test cases.

[Former Member · Accepted solution]

I am not in the HA tests for instance.

I tried rm datafiles (the database continue to survive until restart and need restore after that)

I tried rm log file (the database crash at the first next write and need restore)

I didn't tried the rm of shared nor ini config file

I will try TPC-H test next month !

Do you use XFS or NFS or other FS ?

[Former Member · Accepted solution]

thanks a ton all for your inputs.

Have collated all the HA test cases and executing them so closing the thread.

[Former Member]

unbind the network IP and test it. Killing of sapstartsrv & sapstart will give the same impact of

HDB stop

You can also create some data in master node and check if it is there in slave.

Regards,

Pavan Gunda