> * more complex (two instances failover)

> diff --git a/test/data/htools/hbal-desiredlocation-2.data 
> b/test/data/htools/hbal-desiredlocation-2.data
> new file mode 100644
> index 0000000..a337535
> --- /dev/null
> +++ b/test/data/htools/hbal-desiredlocation-2.data
> @@ -0,0 +1,10 @@
> +group-01|fake-uuid-01|preferred||
> +
> +node-01|16384|0|14236|409600|306600|16|N|fake-uuid-01|1|power:b
> +node-02|16384|0|16384|409600|306600|16|N|fake-uuid-01|1|power:a
> +
> +inst1|1024|51200|1|running|Y|node-01|node-02|drbd|power:a|1
> +inst2|1024|51200|1|running|Y|node-01|node-02|drbd|power:b|1
> +
> +htools:desiredlocation:power
> +htools:nlocation:power

> +./test/hs/hbal --print-nodes=name,pcnt -t 
> $TESTDATA_DIR/hbal-desiredlocation-2.data
> +>>>2/Final cluster status:
> + Name    pcnt
> + node-01    1
> + node-02    1/
> +>>>= 0

Maybe I wasn't clear about this in my last review email. Just counting the 
number
of instances per node in the final configuration cannot, in this example, 
distinguish
between no changes and the correct path. However, when we write tests for a new
feature, we write them in such a way that the test would fail had the feature 
not
been implemented. Here, current hbal (without your changes) would not do 
anything
also resulting in a situation with one instance running on each node, hence the
test would pass as well. So you really do have to check "Solution length" or
a similar indicator to make this test meaningful.

Thanks,
Klaus

-- 
Klaus Aehlig
Google Germany GmbH, Dienerstr. 12, 80331 Muenchen
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg
Geschaeftsfuehrer: Graham Law, Christine Elizabeth Flores

Reply via email to