[torquedev] Proposed Patch

Craig Macdonald craigm at dcs.gla.ac.uk
Tue Mar 16 16:13:59 MDT 2010


> While more internal testing via Adaptive is a good thing,
> it can't possibly match the different uses by the community. For example, would
> AC, thought of testing the qsub -I usage that Brock Palen pointed on on the
> users' list? They figured it "wouldn't break anything" but it turns out
> something did in fact change.
>
> I guess that as long as the community continues to be involved with the test and
> release cycle,
Can I ask what kind of automated testing is done for Torque?

With the advent of virtualising, it should be possible to build a test 
framework, to test as many
things work as possible. Once a suitable framework is in place, its easy 
to require that any new
fix or feature includes a suitable testcase.

Without such a testing framwork, I fear that regressions may be a 
continuing challenge.

Another distributed framework with considerable unit and end-to-end 
testing is Hadoop.
However, it doesn't require root. As Torque does, I guess that 
virtualised images would be
a good place to start.

[Hope this doesn't come across as preaching]

C


More information about the torquedev mailing list