[Linux-ha-dev] Linux-HA CVS: cts by zhenh from

[Lars Ellenberg]

/ 2004-11-08 11:41:02 +0100
\ Lars Marowsky-Bree:
> On 2004-11-08T17:25:47, Huang Zhen <zhenhltc at cn.ibm.com> wrote:
> 
> > Second, adding resources is a part of configuration of the cluster. At 
> > least the original CTS is not do the configuration. A configure tools 
> > will come soon. Some new test cases may be add to CTS to test the 
> > configuration. They will come later.
> 
> I see. Yes, indeed, the CTS as of old did not have the idea of
> automatically configuring the cluster, so this explains this. I agree
> it's probably easier to stick with this model for the first test runs,
> as it's a fairly fundamental change to the CTS paradigm.
> 
> I think that's an important thing to add for the v2 stuff though, as
> testing that the configuration replication works is very essential.
> (As heartbeat didn't use to have it, it wasn't so much of an issue.)
> 
> And again, while this has been mentioned numerous times in the
> requirement discussions on this list, note that we originally planned
> that lge would implement the bigger testing solution, though right now
> I'm not sure how much of it will get done before he departs to New
> Zealand.

unfortunately not much.

I have much of it in bits and pieces, though.  (nothing in python yet) 
mostly worked after the functionality of andrews test shell scripts,
but improved in syntax and line noise as well as usability and
flexibility (all in bash). and the bash part of the CTH (the drbd
test thingy I wrote in perl and bash flavours).
Andrew and lmb have a almost recent version of parts of it.

It "just" needs to be put together, and tested,
and then test cases/scenarios have to be scripted.

my flight leaves early tomorrow, I'll be reading email again next
weekend.  my sister will give me their computer room as guest room, so
while I stay there, I should have plenty of opportunity to hack on this.
I have no idea yet what else will distract me there, but I expect things
to be useable by "not-me"s end of this month.
then it should be in cvs, too.

And indeed, the "automatic configuration" part of whatever test we want
to run with v2 is probably best done with some additional configuration tool,
a number of config templates, and a simple file with a list of templates
to use for this test run.

Yes, this is again sort of a no-not-yet-but-tommorrow-for-sure mail,
and I feel bad about it. unfortunately I did not get too much things
straight in the last few weeks.

	lge