[Linux-HA] configuring CTS - (Was: Wrong includes when using colocations with groups)

Marian Neubert linux-ha at tesla-crew.de
Wed Oct 17 15:10:51 MDT 2007


Dave Blaschke schrieb:

>> BTW: how do i configure the CTS correctly?
> Ideally you'll want to run CTS from a third machine - the test exerciser 
> - that monitors the behavior of the two nodes in the cluster.  The ha.cf 
> and heartbeat version should be the same on all three machines, and 
> syslog-ng should be setup to send messages to the exerciser.
>> when i explicitely enable stonith-testing with the command:
>>
>> python ./CTSlab.py -v2 --stonith yes --standby yes 5 > /tmp/cts_output 
>> 2>&1 &
>>
>> then none of these tests are executed (see attached logfile).
> You did enable stonith tests, but tests are chosen randomly from all 
> runnable tests, so with only 5 tests your chances of running stonith are 
> small.  If you want to run just one type of test, you can use the 
> --choose option to do so (i.e. --choose stonith).
>>
>> stonith is enabled and configured to reboot failed nodes. a 
>> stonith-clone with "apcmastersnmp" is configured, running and working 
>> (tested earlier by hand). startup-fencing is enabled.
> CTS comes with it's own built-in configuration (it generates its own 
> CIB), and the stonith device it uses is ssh.   So even if you have 
> apcmastersnmp setup correctly, it will not be used by CTS.

before writing about this issue i've already set up everything according 
to the cts-README-file. Also the tests are already running BUT i didn't 
know that the cts is for testing the cluster-software only, not my 
previously written cluster-configuration with all its resources and 
constraints...

that's not really clear in the README-file (except the ssh-plugin for 
StonithdTest). Also i didn't know that the tests are run randomly on the 
nodes based on the given iterations.

IIRC the option "--choose" isn't documented in the README - are there 
sources for a complete description what the CTS do in detail and how to 
configure all these details? Is there also a way to test the cluster 
with my own resources configured?

Regards
/Marian


More information about the Linux-HA mailing list